Changelog

​

Enrichment

• Accuracy first: Most enrichments check the sources that test most accurate and complete for that detail first. Then come the ones answering best on recent runs, then the rest, cheapest first.
• Priced per source: Every source has its own price, and a row is charged the price of the one that answered. Misses and reused answers stay free.
• Most and usual cost: Where an enrichment’s sources are priced differently, the catalog, the column editor, run menus and approvals show the most a row can cost and what it usually costs. See What enrichments cost.
• Phone and personal email: Both are priced close to what their sources cost.
• Unused credits released: A run holds the most it could cost and releases what it did not use when it finishes, as a Use AI column does.
• Activity signals and buying intent retired: Find a person’s activity signals and Track company buying intent are gone from the catalog.

​

Find leads

• LinkedIn post and comment signals retired: LinkedIn posts by a person, LinkedIn comments by a person and LinkedIn posts by a company are gone. Find who writes about a topic with People posting about a topic. See People posting about a topic.
• Tables you already have: A table built from one of them keeps its rows, and running its search again still works.

​

People posting about a topic

• And boxes: Add an And box under Phrases, and a post needs a phrase from each box. Search for clay and any of expensive, credits or cost.
• Comments floor: Set Minimum comments to leave out quiet posts, alongside the reactions floor.
• Sharper table names: A new table is named after the first phrase in Phrases and the first in the first And box, so two searches about clay get different names.
• In the AI Assistant: Ask it for posts that mention two things, or for a comments floor. It refuses an edit that would empty a table’s search before you approve it. See Chat.

​

Bug fixes

• Emptied Exclude boxes: An Exclude box from a past or saved search no longer disappears when you delete its last phrase.
• Leftover empty boxes: Opening a past or saved search, or clicking Clear all, no longer keeps empty Exclude boxes opened on the search before.

​

People posting about a topic

• Phrases in and out: Add phrases to include, any of which counts, and phrases that rule a post out.
• A window that moves: The search reads the past month by default, or anything from the past week to any time. It counts back from each run, so a scheduled search reads the latest posts.
• Free count and preview: The count of matching posts and a preview of a few authors cost nothing. You pay for each person added.
• No one twice: Run the search again, by hand or on a schedule, and people already in the table get their newest posts instead of a second row.

​

Find leads

• Grouped by source: More ways to find leads groups its searches and signals by where they find leads: Google, LinkedIn, social media and the web. See People and companies.
• Source requests: Contact us, at the end of the section, opens a short form to tell us which source to add.
• LinkedIn watches removed: The five watches and their pause, resume and stop controls are gone. To find people writing about a topic, use the new search.
• Fourteen signals removed: News coverage, Financial results, Earnings calls, Patent filings, Patent inventors, Website changes and Podcast mentions are gone. So are Podcast appearances, Conference speaking, Glassdoor reviews, SEC 8-K current report, SEC Form D funding, Federal contract awards and New business formations.
• Twelve signals kept: The LinkedIn, X, YouTube, Reddit, GitHub, Hacker News and Product Hunt signals stay. For companies that just raised, use the funding filters under Signals on Find companies.

​

Inbox

• Replies waiting on you: The inbox holds only conversations waiting on you. Answer one, and it moves to Sent. See Inbox.
• Back after three quiet days: An answered conversation that is still open returns after three business days without a reply, tagged No answer in 3 business days.
• Done: Click Done in a conversation’s header, or press E, to clear it until the next message. Select several to mark them done together.
• A reminder choice before you send: Under More reply options, If they do not reply offers Remind me in 3 business days or Do not remind me.
• Kept out after sending: The toast after you send offers Don’t bring back, which keeps the conversation out until the next message.
• In the AI Assistant: Ask it to mark conversations done. See Chat.

​

Sequences

• AI Assistant edits on the canvas: Steps the AI Assistant adds show on the Builder tab at once, marked New. See Build a sequence.
• Undo from the bar: A bar at the foot of the canvas says what the AI Assistant changed, with Undo to take it back.
• Rewritten prompts marked: Review the prompt opens the step with removed words struck through and added words marked.
• Your edits kept apart: Review changes lists only your own edits. What the AI Assistant saved sits under Already live or Already saved.
• Steps placed where you say: On a branching sequence, the AI Assistant asks which path gets new steps, unless you named the step they follow.
• A shorter Summary: The Overview tab’s Summary shows three alerts, with what stops sends first, and folds the rest under Everything else. See A sequence isn’t sending.
• Pause reasons in one place: On the Overview tab, a paused sequence’s reason shows once, in the Summary, instead of again in a banner.

​

AI Assistant

• A new conversation each time: The side panel opens on a new conversation, unless a reply is still being written or a card waits for approval. See Chat.
• Recent conversations: Click the conversation’s name at the top of the panel to switch back to a recent one.
• The edit you asked for: The AI Assistant changes only what you asked and keeps everything else, steps, waits and settings included.
• Whole spreadsheets: It reads every row and sheet of a file you attach, including files with a title row and cell comments.
• On a phone: On a narrow screen the panel is a sheet at the bottom of the page. Once closed, it stays down until you raise it.

​

Tables

• Export from the Tables list: Tick tables and click Export, or choose Export to file from a table’s row menu. Several tables download as one zip. See Export a file.
• Duplicate from the list: A table’s row menu offers Duplicate, and the toast after it offers Open. See Create a table.
• Tables on a phone: Lists and panels fit a phone screen, and a long press on a cell opens its menu.
• Runs over every row: On a larger table, the arrow beside Create & run 10 rows now offers a run over every row. See Enrichment columns.

​

Find leads

• More companies matched: A company website in a people search also matches its main domain and the site it redirects to, so a followed table finds more people. See Saved searches.
• Companies that match no one: A warning counts companies with no people found, and Companies that match no one lists them with same-name suggestions.
• Pasted job titles: The bulk add list on Job titles keeps every title you paste, including ones outside the suggestions.
• LinkedIn watches moved: The five watches now sit under Watch LinkedIn, inside More ways to find leads.

​

Enrichment

• Second check for catch-alls: On a Verify email column, switch on Check catch-alls with a second verifier to send catch-alls to another verifier. A row is charged the second check only when it is settled. See Email and phone.
• Low confidence emails: An email at a different domain from the person’s company is marked Low confidence, so you check it before sending.
• No guessed surnames: A person whose last name shows only as an initial is never looked up by name. With nothing else to go on, the row is skipped and nothing is charged.

​

Bug fixes

• Moves to Trash before a reload: A table moved to Trash moments before a reload now stays in Trash.
• Answers kept during edits: Editing an unrelated cell no longer drops answers still being filled in by a running column.
• Stale credit status: A table’s “Paused: out of credits” status no longer stays after another run has filled the rows it was waiting on.
• People dropped from searches: A people search no longer leaves out people it matched when the same person came back twice.
• Hiding the checklist: Hide checklist now closes the setup checklist, instead of leaving it open under the toast.
• Blank sign-in page: When sign-in fails to load, the page says so and offers Try again instead of staying blank.

​

AI apps

• Sign-in instead of keys: An app signs in as you and works in the one workspace you pick. API keys no longer connect AI apps, and still work for the REST API.
• Read only or read and write: Choose what the app may do when you sign in. Making changes from an AI app needs a paid plan.
• The app’s prompt is the approval: Before a change, the app asks you to allow the tool. A change that costs credits says what it was priced at.
• Connected apps: AI apps in Settings lists every app connected to the workspace. Disconnect your own, or anyone’s as a workspace admin.

​

Find leads

• Find Google results: Paste the searches you would type into Google and get a table of the results Google shows. Each row is one result, with the search that found it and its rank. See Google results.
• Searches as written: Operators such as site: and intitle: work, and cost the same as a plain search.
• Depth per search: Pick how many results each search adds. A search stops sooner when Google runs out.
• Country: Pick the country the searches run from, and Google ranks the results for searchers there.
• Pay for what lands: You pay only for results added to the table, and the preview is free.

​

LinkedIn watches

• Five things to watch: Posts you paste, every post from people or company pages you name, posts about a topic, and LinkedIn events.
• Who gets added: Choose who around a post counts, such as the people who reacted, commented or wrote it. Then narrow by role, place and company.
• A daily limit: Set how many people a watch adds a day, and an optional total limit. You pay for each person added, and nothing before that.
• Enrichment as people arrive: The enrichment you pick when you start runs on each new person as they are added.
• Status on the table: A banner says what the watch is doing and how to fix a pause. The table’s source panel pauses, changes or stops it.
• No one twice: A watch never adds or charges the same person twice. Starting it again never charges for people already in the table.
• In the AI Assistant: Ask it to start, pause, resume, stop or change a watch. Starting, resuming or raising what it may spend waits for your approval. See Chat.

​

Sequences

• Steps added after the end: A step added where contacts finished reaches them too. Each wait counts from their last step. See Build a sequence.
• Deleted steps carry contacts on: Delete the step contacts wait for, and they go on to the next step or the branch you keep.
• Approval changes follow waiting contacts: Turn approval off, and steps only it held go out. Turn it back on, and what it released that has not gone yet returns to Tasks.
• Rewrites after writing changes: Change a step’s prompt or how the agent writes, and its unsent messages are written again. Edited or approved ones stay. See The outbound agent.
• Window changes move waiting contacts: A wider sending window brings waiting contacts forward, and a narrower one moves them later. See Schedule and sending window.
• Contacts back after the cooldown: Once the window passes, contacts the cooldown refused before anything reached them start at step one. See Cooldown.
• Removed senders: Removing a sender stops the contacts its mailboxes already emailed at their next email. Contacts it never reached go on with another sender.
• Writing as messages send: The agent keeps a limited number of contacts written ahead on each channel. The rest read Queued to write until a place frees up.
• Every mailbox takes its turn: A contact’s first email goes from whichever of their sender’s mailboxes can send soonest, inside the sending window. See Senders.
• Senders in Settings: A sequence’s senders are now the first section of its Settings tab. See Build a sequence.

​

Tasks

• One card per email thread: The agent writes a contact’s follow-ups with their first email. Edit any of them, then approve or reject the whole thread. See Tasks.
• After-hours approvals first: An email approved outside the sending window goes out first when the window opens.
• Approvals by sender: Approve tasks on a sender’s LinkedIn tab or in the side panel now opens only that sender’s LinkedIn tasks.

​

Inbox

• Replies only: The inbox lists only conversations the contact answered. A sequence email nobody replied to stays out until they do. See Inbox.
• Replied filter: Set Replied in Filters to see conversations you answered, or Not replied for those still waiting on you.
• Resizable list: Drag the line between the list and the conversation to resize them. Double-click it to reset.
• LinkedIn replies skip Tasks: A LinkedIn reply you send from the Inbox never waits in Tasks, whatever the sender’s run mode.

​

Tables

• Answers kept for unchanged inputs: When a row changes, a paid column keeps its answer if what it reads is the same, and nothing is charged. See Re-runs and cached results.
• Refreshes follow run settings: After a refresh from a connected tool, columns fill the rows it added or changed, plus any still out of date, and only while Run automatically on new and changed rows is on. See Table settings.

​

Notifications

• What each alert covers: Each notification type now has a short line saying what happened. See Notifications.

​

Bug fixes

• Running elsewhere after a reconnect: Reconnecting a browser no longer leaves the side panel reading Running elsewhere when no other browser holds the sender.
• Duplicate steps from quick clicks: Queuing a step from Try a step several times at once now queues one.
• Withdrawals and capital letters: A withdrawal from Try a step now finds the pending request when the profile address uses different capital letters.
• Limit set to 0 in the side panel: A step held by a limit set to 0 now says so in the side panel, instead of Ready.
• Skipped steps in the side panel: A step you skipped reads Skipped in the side panel’s history, instead of Rejected.

​

Tasks

• Quieter approvals: Approving a task no longer raises a message over the next task’s buttons. A failed approval still says so. See Tasks.
• Following approved work: An approved row leaves Needs approval. Pick Approved under Filters to see where each row stands. See Tasks.

​

LinkedIn

• Requests after a withdrawal: LinkedIn refuses a new connection request for a while after one is withdrawn. The failed step now says so and when to retry.

​

Check the writing

• One box for notes: What should change sits under the whole thread. It replaces the corrections on each message and the separate note for all messages.
• Rewrites in place: Each message shows its earlier words above the new ones, with the lesson in one line. Looks good keeps it, another note refines it, and Discard the rewrite drops it.
• The next contact: Looks good moves on to the next contact in Writing for that nobody has judged. The one after it is written ahead, so you do not wait.
• Send labels: Each message reads Sends on its own or Waits for approval, with the reason it would wait. See Tasks.
• Training progress: A count above the thread shows how many threads you reviewed and fixed. It says when the agent is ready to send on its own, and What it learned lists each lesson.
• Approvals kept: Threads marked Looks good keep their words when the agent learns a lesson. Before, they were written again at send.
• In Chat: Ask Chat to rewrite a contact’s thread from your note, keep or discard the rewrite, and move to the next contact. See Chat.

​

Simpler campaign tailoring

• Prompts written for your company: Before you create the sequence, you see the campaign, why it fits, the goal and every step’s prompt.
• Goal set for you: Tailoring picks one of the goal presets. Change it before you create the sequence, and a goal you chose is never replaced.
• Drafts only: Tailor again works on drafts and rewrites every step’s prompt. A launched sequence keeps its steps.
• Fewer steps to finish: The check against the standard campaign and the review on the campaign card are gone. A finished tailoring goes straight onto your draft.

​

Sender activity

• Each row’s decision: A row says whether a person or an API key approved it, it went automatically, or someone wrote it in the Inbox.
• Filters and sorting: Narrow by channel, decision, outcome or When, search by contact or subject, and sort by date or contact.
• LinkedIn history on Activity: The LinkedIn tab no longer has a history table. Steps that reach someone are on the Activity tab, and the extension’s side panel lists every step, checks included. See Connect LinkedIn.
• Running step on the queue line: The LinkedIn tab’s queue line names the step a browser is running now.
• Activity in the AI Assistant: Ask what a sender sent and who approved it. The AI Assistant reads the same list. See Chat.

​

Sequences

• Choose which branch carries on: Deleting a condition or connection request with steps on more than one branch asks which branch to keep. See Branch on a condition.
• Misplaced withdraw check: A withdraw left on a request’s Connected branch is flagged at launch. Its panel moves it to Not accepted in one click. See Launch checks.
• Shorter acceptance waits: Shorten how long a connection request waits, and contacts already past it go down Not accepted. The review before you apply counts them. See Build a sequence.

​

LinkedIn

• New connection checks: Oneprofile reads each sender’s newest connections to spot accepted requests. These checks run at any hour, in either run mode. See Connect LinkedIn.
• Connection check limit: Set a daily limit for Check new connections under Housekeeping on the sender’s LinkedIn tab. The tab warns when the checks fail or are switched off.
• One profile look when unsure: When a wait ends with no answer seen, Oneprofile looks at the person’s profile once. It waits in Tasks whenever the sequence’s LinkedIn steps do. See Tasks.
• Accepted requests in analytics: The LinkedIn senders table counts accepted connection requests beside their share of requests sent. See Sequence analytics.

​

Bug fixes

• Steps after a laptop sleeps: A LinkedIn step no longer fails because the sender’s laptop slept or Chrome froze mid-step. It finishes once the browser wakes.
• Redrawn LinkedIn buttons: Steps no longer fail when LinkedIn redraws a button just as it is pressed. A step that loses its button doesn’t count against the daily limit.
• Pending requests behind More: A request already pending behind LinkedIn’s More menu now reads as pending, not as Connect not offered.
• Follow, like, message and withdraw: Each step now finds LinkedIn’s current controls, including those behind the More menu, so fewer of them fail.
• Retrying with a contact taken elsewhere: Retrying every interrupted contact no longer fails when one is live in another sequence. That contact stays failed and is counted.
• Profiles LinkedIn can’t find: Once LinkedIn says a profile doesn’t exist, later LinkedIn steps for that address are skipped. A corrected address is tried again.

​

LinkedIn steps that carry on after a closed tab or a restart

• Closed tab: The extension reopens its LinkedIn tab in the background and finishes the step. Closing it twice within ten minutes pauses steps in that browser until you click Resume in the side panel.
• Restarts and lost connections: A step running when Chrome or the extension restarts, or when the connection drops, finishes once the browser is back.
• Retry checks first: Retry on a failed LinkedIn step checks LinkedIn first. A step it finds already done is marked done and charged once.
• Already pending: A connection request that was pending before Oneprofile sent one reads Already pending, and it isn’t charged.
• Sent after all: A step marked failed that later turns out to have gone through says so in its detail.

​

Bug fixes

• Steps failed while still running: A step could be marked as not finishing in time at the moment its browser checked in. It now finishes.
• Two LinkedIn steps at once: Two quick check-ins from one browser could start two steps together. A sender now runs one LinkedIn step at a time.
• Daily limits after a timezone change: A step that didn’t run gave its slot back to the wrong day once the sender’s timezone changed. It now goes back to its own day.
• Results lost on a restart: A step whose result was still saving when the extension restarted could end up marked failed. The result is now kept and sent again.

​

Connection requests that branch on the answer

• Branching connection requests: Every connection request has a Connected branch and a Not accepted branch. Set how long it waits on the Not accepted branch, and a wait after Connected counts from when they accepted.
• Default withdraw: A new connection request comes with a Withdraw connection request on Not accepted, so a request nobody answered is taken back. A withdraw can’t go on Connected, where there is nothing to take back.
• Already connected branch: Contacts already connected to the sender go down Connected with everyone who accepts. To send them their own way, switch on Give already-connected contacts their own branch under Advanced.
• Existing sequences: Connection requests already in a sequence branch too, with both branches leading to the step that followed, so that step now waits for the answer. Connection request accepted? left the step picker, and a check right after a request is now part of it.
• Reasons in contact history: A contact’s history says why they took a branch, such as “LinkedIn didn’t offer Connect on their profile”. A contact the request can’t reach goes down Not accepted rather than on to the next step.
• AI Assistant: Ask it to put steps on either branch, give already-connected contacts their own branch, or change how long a request waits. See Chat.

​

Bug fixes

• Already-connected contacts on the wrong path: A contact already connected to the sender got no request, so a Connection request accepted? check after it answered No. They now go down Connected.
• Contacts stuck on an acceptance check: A contact could wait on Connection request accepted? long after its window ended. Once the window ends with no answer, a look at their profile now settles it.

​

Senders and mailboxes

• Sender timezones: Each sender has a Timezone, detected from the browser that connects their LinkedIn, or else from whoever added them. Change or clear it on the sender’s Settings tab, or ask the AI Assistant. See Senders.
• Limits reset at the sender’s midnight: From October 3, every daily email and LinkedIn limit resets at midnight in the sender’s timezone, not midnight UTC. Weekly LinkedIn limits count the sender’s last seven days.
• Timezone changes keep the day: When a sender’s timezone changes, the day already under way finishes first, so a change never starts a fresh day early. Workspace analytics and charts still count days in UTC.

​

Sequences

• Hours in each sender’s timezone: A sequence’s sending hours now apply in each sender’s own timezone. A New York sender and a London sender each send 9-5 their own time. See Schedule and sending window.
• Existing sequences switched too: Every existing sequence now sends in each sender’s timezone. If its hours were meant for your prospects’ timezone, pick that zone under Timezone in its Schedule to keep it.
• One sender per contact: Each contact gets one sender at their first step and keeps them, so all their steps go out in that sender’s hours.

​

LinkedIn

• LinkedIn steps keep to sending hours: Every running sequence’s LinkedIn steps now stop outside its sending hours, in the sender’s timezone. Before, a queued step could run at any hour. See LinkedIn steps.
• Outside sending window: A step waiting for the hours reads “Window opens in 3 hours” in Tasks. The sender’s LinkedIn tab and the side panel say so too. Checks of your own inbox and sent requests still run at any hour.
• Slower default pace: A new sender now waits longer between LinkedIn steps, so a day’s steps spread out the way a person browses. Senders that never changed their Time between steps moved to the new default too. See LinkedIn steps.

​

AI Assistant

• Timezones by chat: Ask it to set or clear a sender’s timezone, or to hold a sequence to one timezone’s hours. See Chat.

​

Sequences

• Contacts start at launch: After Launch, contacts due now start straight away, rather than up to a minute later. A message follows them until they have started or their steps wait for approval. See Build a sequence.
• Resume and retry start at once: Resuming a sequence or a contact, or retrying one, starts whoever is due straight away too.
• Send now: Send now in a contact’s row menu, or on a selection, makes their next step due at once. The sending window, limits and approval still apply. See Choose the audience.
• Audience order follows the view: Contacts are reached in the order the view shows them at launch. A view sorted by score reaches your best leads first.
• Edited waits reach waiting contacts: Contacts already in a wait take its new length, counted from when their wait began.
• Resumed contacts keep their timing: A resumed contact gets their next step when their wait would have ended, not the moment you resume them.

​

Tasks

• A live list: Tasks updates on its own, with no reload, and so does its count in the sidebar. A sequence’s Audience tab and a sender’s LinkedIn tab do too. See Tasks.
• Where approved work stands: A Detail column says where each approved row stands and counts down to it. View approved, in the message after you approve, opens that list.
• Failed steps: Failed lists the steps that failed, newest first. Open one to read why, and Retry runs it again.

​

LinkedIn

• Why the next step waits: The sender’s LinkedIn tab and the extension’s side panel say why the next step waits and until when. See Connect LinkedIn.
• One failure stops one contact: A failed LinkedIn step stops only that contact, who reads Interrupted until you retry. The rest of the audience keeps going. See LinkedIn steps.

​

AI Assistant

• When work goes out: Ask when approved steps will run, or why a sender’s LinkedIn is waiting. It also says how many contacts are still starting after a launch. See Chat.
• Failed steps by chat: Ask which LinkedIn steps failed and why, then have it retry them.

​

Bug fixes

• LinkedIn steps starting late: The extension could miss the end of the time between steps and start the next step well after it. It now checks again as the gap ends.
• Emails pushed to the next day: With several contacts due on one mailbox, only the first could send and the rest kept moving later. Each now sends when its slot comes.
• LinkedIn-only sequences at the table limit: A sequence with no email step could not launch while the workspace was at its plan’s table limit. It launches now.

​

AI Assistant

• Before and after on every edit: An edit’s card lists each change. Show what changes sets what it says now beside what it will say.
• Changes that always ask: Sending, removing, stopping contacts, adding contacts to a running sequence and loosening what waits for a person always wait for Approve. So do letting an agent act without asking, rewriting your business context from your website, and disconnecting a tool. See AI Assistant.
• Older cards closed: A card raised before this update that would now do something different was closed with a note. Ask again for a new card.

​

Tables

• Rows that match either condition: Ask for rows where industry is SaaS or Fintech, with bracketed groups for anything more involved. See AI Assistant.
• Several columns at once: The assistant runs several columns in one go, on the rows you picked.

​

Find leads

• Saved searches run again: Ask the assistant to run a saved search, or one from the search history. Counting is free, and adding rows to a table asks first. See AI Assistant.

​

Sequences

• Tailored campaigns by chat: Accept, dismiss or reset a tailored campaign from the conversation.
• Stop, pause or remove by filter: Name who to act on, such as contacts who bounced, opted out or are out of office.
• Contact history and eligibility: Ask for one contact’s journey through a sequence, or why someone is not being contacted. See A sequence is not sending.

​

Senders and mailboxes

• The LinkedIn queue in Chat: See what waits in a sender’s LinkedIn queue, then approve, reject or cancel steps in it.

​

Agents

• Duplicate agents: Ask the assistant for a copy of an agent, or to save it as a template.

​

Bug fixes

• Edits that changed more than you asked: An assistant edit could switch off or reset a setting nobody mentioned. Everything you do not name now stays as it is.

​

Sequences

• Move steps: Drag the grip beside a step’s name onto Move here on any connector, and it moves with every step after it. Undo puts it back. See Build a sequence.

​

Chat

• Every lead filter: Chat can set any filter the Find leads page offers, on a new search or on the search behind a table. See AI Assistant.
• Changing a table’s search: Asking Chat to change the search behind a table changes only what you name, and keeps every other filter.

​

Bug fixes

• Branches lost in a rewrite: When Chat rewrote a sequence’s steps, a branch could come loose, and steps it kept could lose their words and waits. Kept steps now keep where they go and everything you did not ask to change, and the card lists each change before you approve it.

​

Enrichment

• More work email sources: Find work email checks another source when the first ones miss, so more people come back with an address. See Email and phone.
• Phone numbers from a name: Find phone number reaches more sources, and more of them work from a name and company when there is no LinkedIn URL.
• More technology stacks: Find technology stack checks another source before its last resort, so more companies come back with their technologies.

​

Bug fixes

• Switchboard numbers as a person’s phone: Enrich person from email could fill the phone with a company switchboard, branch line or fax. It keeps only the person’s own line now.

​

Sequences

• Tailored campaigns: Click Tailor a campaign, pick a campaign or Best fit, then a table and senders. The agent tailors it to your audience before Create sequence makes the draft. See The outbound agent.
• A check against the standard: Each tailored campaign is tested against the standard one on a few contacts. When the standard reads better, it is kept, and the card says why.
• Campaigns on the Sequences page: Start with a campaign sits above your sequences. The New menu offers a campaign, a saved template or a blank sequence. See Build a sequence.
• Tailored from a table: Sending a table to a sequence now starts on Tailor a campaign, with a sender picker beside it.
• Missing work emails: Launch checks offer to find work emails for contacts who have none. The cost shows before you confirm. See Launch checks.
• Unreachable contacts: A contact with no address for the sequence’s channels reads Unreachable, and the tooltip names what is missing. See A sequence is not sending.
• Your only sender, added for you: A workspace with exactly one sender gets it on every new sequence. See Senders.

​

The outbound agent

• The Writing panel: It opens on who writes which email. Hand the first email to the agent, or take every email back with Write all yourself. See The outbound agent.
• One contact’s whole sequence: Check the writing shows every message for one contact in order, with the waits between them. Mark each Looks good or Needs work.
• Rewrite from feedback: Pick what was wrong with a draft, and Rewrite with this updates the step’s instructions and writes it again. The earlier draft shows beneath for comparison.
• Write from your columns: The Write from row lists your research and AI columns, which the agent uses as angles. Untick any you want left out.
• Columns in a prompt: Type / in a prompt or goal to point at a column. Launch checks flag a column the audience does not have. See Email steps.
• Voice: Business context has a new Voice section for spelling, formality and wording. Every outbound email and reply draft follows it. See Business context.
• Written as the sender: The agent writes with the sender’s job title, and each contact keeps one sender for the whole thread.
• Research on from the start: New sequences start with Web research and Person research switched on. Switch either off to spend less.
• Contacts it cannot confirm: When research cannot confirm who a contact is, the first message is not written. The preview says what to check in the row.

​

Tables

• Answers row by row: Each cell’s answer appears as soon as it finishes. Columns built on it fill in on the same row at once. See Run columns.
• Stopping costs nothing extra: Stopping a run keeps the answers already in and drops cells still working, without charging for them.
• Charged as it runs: A column’s spend updates while the run works, and one run shows as several lines on the Usage page. See Credits and billing.
• Up to, and usually: Run menus, confirms and AI Assistant cards quote the most a run can cost and what it usually costs. Cards show the real charge afterwards.
• One progress meter: The footer shows one percentage done, counts cells no run has reached, and links to the failed rows.
• Conditions in the formula builder: The conditions step has its own editor, and a condition can compare against an earlier step’s result. See Function columns.
• Failed cells in plain words: Hover a failed cell to read the reason and whether anything was charged.

​

Enrichment and research

• Company contact emails: Find company contact emails checks common inboxes such as info@ and contact@ first, so it returns more of them. See Email and phone.
• Deeper LinkedIn profiles: Reading a LinkedIn profile adds skills and recent posts, and headlines and work history come back more often. The cost shows before you run it.
• More matches: Technology stack, phone number, personal email and email verification lookups reach more sources.
• Partial answers kept: When one field of a multi-field AI answer cannot be found, the fields that were found stay in the cell.
• Whole pages read: Agents now read a page’s header and footer, where addresses, phone numbers and contact links usually sit.
• Files spotted first: A link to a PDF or other file is recognised before it is read, and the cell says to use a web page. See Web research and context.
• Business context at signup: A new workspace’s business context is drafted from the company website on first load. Anything you type meanwhile is kept.

​

AI Assistant

• Run reports in the chat: When a run started from Chat finishes, the conversation says what filled in. One tap reruns the empty or failed rows. See Chat.
• More people from the same search: Ask for more people on a table built from a lead search, and the assistant reaches further into that saved search.
• Campaigns by chat: The assistant can tailor a campaign and find missing work emails, each on its own approval card.
• An empty business context: Before adding columns that write, the assistant offers to write your business context from your website.

​

Bug fixes

• Mailboxes marked disconnected: A brief network or Google outage could mark a Gmail mailbox disconnected and email its owner. Only a real refusal disconnects it now.
• Replies missing from the Inbox: A message Gmail would not return during a sync was skipped for good. It is retried now, so the reply shows and stops its sequence.
• Progress that read done too early: Running a few rows of a large table showed the footer at 100%. It measures the whole table now.
• Answers lost in a failed batch: A batch that failed or timed out threw away the answers it had already written. Finished answers are kept now.
• Raw errors on failed cells: A failed AI cell could show the model’s internal error. It gives a plain reason now.
• Check the writing greyed out: Check the writing was disabled with no audience, or with changes waiting. It opens the audience chooser or the changes review instead.
• Escape and the cell panel: Escape cleared the selection but left the cell panel open. It closes the panel now.
• Raw error pages: Some screens, including sign-in, could show a raw error page. Every screen now has a plain one with a refresh button.
• Messages held for a shared first name: A contact who shared the sender’s first name had every message held for a person. Those messages can send on their own now.
• Formulas that lost settings: Re-saving a formula could drop settings the builder had no control for. They are kept now.

​

Tables

• Value distribution: Open a column’s header menu and choose Value distribution. It lists the most common values with counts and percentages, plus distinct, empty and Other values. See Rows and cells.
• Counts while you map a file: Each column in the import mapping shows how many distinct and empty values it holds. Click the counts to open that column. See Import a file.
• Raw values, page by page: A column’s dialog carries a Raw values tab that pages through the file in order, keeping original spelling, spacing and leading zeros.

​

Find leads

• Enrichment is optional: A buying-signal, similar-company, hiring or backlinks search can add rows with nothing ticked. Tick only what you want filled in. See Choose what to enrich.
• Bigger searches everywhere: Every finder can now ask for as many rows as your plan’s tables hold. The first rows arrive with the table, and the rest are added in the background.
• What the rows cost, up front: The price line under Save to a new table names what retrieving the rows costs, beside the enrichments you tick.
• Resizable preview columns: Drag a preview column’s header edge to widen it, double-click to fit it to its values, or press Escape to put it back.
• The page each link points at: A backlinks sample shows the linked page beside the referring site, page title and anchor text. See Backlinks.
• Counting only where it is free: A map or backlinks search no longer counts matches as you type, because a count is not free on these. Choose how many rows to add instead.

​

Agents

• Research that gets past a wall: A page that refuses a plain read is retried by another route, then by a rendered browser. The trail says which one answered. See Web research and context.
• A clock of its own for research: A cell that researches the web has its own time limit, and stops reading in time to write its answer.
• A trail that survives a timeout: A research cell that runs out of time keeps every step it recorded, and says it ran out of time. See Read an agent run.

​

The outbound agent

• A contact’s messages written together: When the first message is due, the agent drafts the rest of that contact’s planned messages in the same pass. Each one sees the messages before it. See The outbound agent.
• One round of research per contact: What the agent finds out for the first message carries the follow-ups, so the same page is not read twice.
• Edits rewrite what follows: Editing or regenerating a message drops the later drafts written after it, so the follow-ups answer your version.
• Training one contact at a time: Write the emails walks contact by contact, with every message for that contact drafted together. The next contact starts after your review.
• Training reads the web too: A training contact gets the web and profile research the agent is set up for, so what you review is what would send.

​

AI Assistant

• Distributions on request: Ask which values are most common in a column, or how many cells are empty, and the assistant reads the same distribution you see. See Chat.
• A file read before it becomes a table: The assistant can inspect one column of an attached CSV or Excel file, both its distribution and its raw values. See Import a file.

​

Bug fixes

• A heading with nothing under it: Opening a table drew Waiting for your approval over a placeholder while the list loaded. The heading waits for a real card now.
• Long replies that dropped: A quiet stretch while the assistant worked could drop the connection before the answer arrived. The connection is held open until the reply lands.
• An error after the answer arrived: A dropped connection showed an error even when the reply had been written. The conversation is trusted instead.
• Columns you could not open: In the import mapping, a column whose read rows were all blank had its expand control disabled. Every column opens now.

​

Find leads

• Backlinks: Name a domain or a page and get the sites that link to it, one row per referring site, best links first. Filter on the referring site, the page, the link and the page it points at, then enrich the rows like any company table. See Backlinks.
• Big searches keep adding rows: A people or company search can ask for thousands of rows. The first rows arrive with the table, and the rest are added in the background. See People and companies.
• One ceiling, not two: The most one search adds is what your plan’s tables hold. Nothing else caps it now.
• Run history reads right: A search still adding rows reads as running, not failed, and says how many rows are in so far.
• People to add, beside the button: Editing a table’s search, the number of rows to add stands beside Save & run search. It used to sit in a closed group in the rail.
• One run at a time: A search that is still adding rows cannot be run again until it finishes, so the same rows are not added twice.

​

Tables

• Progress while a search fills a table: A table still filling from a search carries a banner with the rows added so far. It says when the run finished, and anyone who opens the table sees it.

​

Sequences

• Pick the email a step replies to: Every email step has a Thread control listing the emails before it. Pick one, and the step goes out in that email’s thread. See Email steps.
• Start a new thread later on: Starts a new thread gives a step its own subject, so a sequence can open a second conversation after the first. A blank subject no longer means a reply.
• Subjects written per contact: On an agent-written email that opens a thread, the subject line is the agent’s to write, fresh for every contact. See The outbound agent.
• Cards name the thread: A step’s card reads the subject it opens with, or the thread it replies into, instead of describing the mechanism.
• Launch checks follow the thread: The checklist names any email that opens a thread with no subject, not only the first one. See Launch checks.

​

AI Assistant

• A dock you can resize: Drag the seam beside the docked AI Assistant to make it wider or narrower. Double-click to put it back, or use the arrow keys. The width is remembered.
• Back where you left it: The dock reopens on the conversation you last read in this workspace, whether you read it there or on the Chat page.
• Conversations stay in view: On a wide window the Chat page opens with the conversation list beside it, and remembers when you shut it. See all conversations opens it at any width. See Chat.
• A file bigger than a table: A file with more rows than a table holds is cut short. The assistant now says how many rows went in and how many the file had. See Import a file.

​

Bug fixes

• Searches that repeated a profile: A search whose source returned the same profile twice failed outright. The profile is added once and the search finishes.
• Replies from the wrong mailbox: A follow-up could go out from a mailbox that never opened the thread, breaking the conversation for the reader. A reply now sends from the mailbox that sent the message it answers.
• Follow-ups that lost the thread: A reply carries the whole chain of the conversation, so Gmail and Outlook file it under the opening email.
• The AI Assistant restarted on every page: Moving between pages threw away the reply it was writing and your place in the conversation. It stays open as you move now.
• A preview the agent could not write: When the agent cannot be reached, the preview says it is temporarily unavailable instead of repeating the model’s own wording.
• Paragraphs in a prompt: A blank line between two paragraphs of a sequence prompt shows as one. The paragraphs used to run together.

​

The outbound agent

• Person research: The agent can read a contact’s LinkedIn profile before the first message and keep what it finds for the rest of the conversation. It runs once per contact alongside web research, priced the same way and only charged when it reads something. See The outbound agent.

​

The outbound agent

• The Agent tab: A new tab on every sequence holds the per-message prompts, the voice, the goal and the call to action, the model, and what the agent knows. See The outbound agent.
• Prompts, not templates: An agent-written step carries a prompt saying what the message is for. The agent writes the opener’s subject and keeps follow-ups in the same thread.
• Preview a real contact: Once a sequence has an audience, preview any step against a real row and read the message that contact would get, with the agent’s confidence and a quality score beside it.
• Research, once per contact: Switched on, the agent reads the web about a person and their company before the first message, and reuses what it found for the rest of the conversation.
• LinkedIn too: Connection notes, messages and InMails can be written by the agent in the same voice.
• It learns: Rewrites on approval cards, thumbs and comments become examples, and the agent proposes rewritten prompts you accept or refuse.

​

Replies

• The agent answers replies: A contact’s reply is answered in the sequence’s own voice. Whether that answer sends on its own is the sender’s send mode; anything doubtful waits for a person, with the reason on the draft. See Reply with AI.
• Marked and notified: A reply the agent sent is marked Sent by the agent in the thread, and you are told each time one goes.
• Two switches: Agent replies stops the writing; Pause agent sending keeps the writing and holds everything for a person. Anyone can pause, only an admin can resume.

​

Senders and approvals

• Send mode: A sender is on Send when confident, Always ask or Always send. New senders send when confident. Always send is an admin’s choice and asks for confirmation. A sequence follows the strictest of its senders.
• Job title and booking link: Both live on the sender, and the agent uses them instead of guessing.
• Tasks says why: The approval queue shows the agent’s confidence, the quality score and the reason a message waited, sorts by score, and offers Write again with a note.

​

Retired

• The starter template gallery, the goal box that generated a whole campaign, and AI snippets inside a message are gone; the agent writes the message instead. Saved message templates are unchanged.

​

AI Assistant

• One history: Every conversation sits in one list, on your account and private to you. See all opens the page that searches, pins and renames them. See Chat.
• Search builder retired: The conversational search builder on Find leads is gone, with its own page and the panel beside the filters. The box on Find leads hands your words to the docked AI Assistant, which runs the search and says how many match. See AI Assistant.
• Table panel retired: The table’s own AI Assistant panel is gone, and so are the column cards it suggested. Ask for the column you want and it adds it, then prices the run on a card. See AI Assistant.
• Agent rail retired: The agent editor’s chat rail is gone. Ask for a change and the card shows the whole revised prompt and exactly what changed before anything is saved. Every save keeps a version. See Build an agent.
• Old links still work: A bookmark to the old search builder opens the Find leads page.

​

Tables

• Excel import: Import a table from an Excel workbook as well as a CSV. Numbers, dates and checkboxes arrive as their own types. See Import a file.
• Sheet picker: A workbook with several sheets asks which one to read, and re-reads the upload when you switch. Each sheet brings its own columns.
• Excel export: Export to file offers Export to CSV (.csv) and Export to Microsoft Excel (.xlsx), from the table’s menu, the Tools panel, and a row selection. See Export a file.
• Types that survive: An Excel export keeps numbers summable and dates sortable, so a downloaded table is ready to work in.
• Older workbooks: An .xls file is refused with the fix, rather than a parser error. Save it as .xlsx and import that.

​

Tasks

• Tasks as a table: Everything waiting on a person sits in one table you can tick, filter and decide on together. See Tasks.
• Decisions on many rows: Tick the rows, then approve them in one go, or reject them with one reason kept on each.
• Rows that moved on: A bulk decision says how many rows it settled, and how many had already moved on while you read them.
• Filters on every column: Narrow the table by status, how long a row has waited, outreach type, Task type, Assigned to and sequence.
• Waiting since: Each row says how long it has been in the queue, in place of a countdown to a time nothing runs at.
• Task type on the row: Every row names its own step, so the list reads as one vocabulary instead of repeating the channel.

​

Approvals

• The sender decides: Whether a channel goes out unwatched is set on the sender’s own page, and nowhere above it. See Senders.
• One place to read it: The workspace approval card at the top of Tasks is gone, so no second screen can contradict the sender.
• A sequence asks, never allows: LinkedIn approval and Email approval now offer Sender default and Require approval. Neither can let a step send unwatched.
• A sender nobody has set: LinkedIn steps wait for a person and email sends on its own, until the sender’s page says otherwise.
• Run mode beside the limits: Each sender’s Manual run or Automatic run is chosen next to Limits on their LinkedIn tab. See Connect LinkedIn.

​

Sequences

• Addresses that arrive later: Type an address into a row you added by hand, and the sequence enrols that contact. See Choose the audience.
• Binding a table says so: Pointing a sequence at a table now says it is reading the table and enrolling rows, rather than looking untouched.

​

Bug fixes

• Launch checks: Contacts already enrolled in the sequence you are launching read as ready to go, instead of reading as left out.
• Unsaved sequence edits: A save that fails keeps your edits on the canvas, with the button that ships them still there.

​

AI Assistant

• Lead search in chat: Describe who you want and it searches for people or companies, then says how many match before anything is added. See Chat.
• Tables from a search: It creates the table, adds the rows, and answers questions about what is in them afterwards. See People and companies.
• Columns it adds and runs: AI columns, enrichments such as work emails, and the run itself, with the price on the card before it starts. See Run columns.
• Sequences end to end: It writes the steps, picks the audience and the sender, and launches the sequence once you approve. See Build a sequence.
• Answers from the help: It searches the same help pages you are reading now, and quotes what something costs before it runs.
• Everything else through the API: A job with no shortcut of its own is reached through the API, so the assistant is never a dead end.
• Plans, one question, next steps: It writes a short plan before a long job. If your request could mean two things, it asks one question with answers to tap, then suggests what to do next.
• Memory retired: It no longer keeps notes between conversations. What it knows about your company is your business context, which you can read and edit. See Business context.

​

AI Assistant

• Undo removed from action cards: A change the assistant made is an ordinary change afterwards, so you take it back where you would take back your own: delete the column from its header menu, remove the contacts from the sequence, move the table to Trash. Approve is still the only thing that spends, and the card still names the price before you press it. See AI Assistant.
• Activity still names the card: A line in a table’s Activity or a sequence’s Activity still names the card the change came from, so what happened and what authorised it read together.

​

Find leads

• Technology search folded in: Companies by technology is no longer a search of its own. Find companies has a Technology section that asks the same question, and it composes with every other company filter. See People and companies.
• Still using it, and just started using it: The Technology section adds two dates the old search never had. Still using it since drops companies that ran a product once and stopped, and First started using it since finds recent adopters.
• Old links still work: A bookmark to the technology search opens Find companies.
• The AI Assistant builds it: Ask for companies using a tool and it sets the Technology section instead of pointing you at another search.

​

Tables

• Chart view retired: Tables no longer switch to a chart. Use a saved view, a filter, or the coverage strip on a column to read the list. See Views.

​

AI Assistant on Find leads

• Questions with answers to tap: A vague request, or a word with two meanings, gets one question with answers to tap. After two questions, it goes with its best reading and says what it assumed. See AI Assistant.
• Steps as it works: Each step appears the moment it finishes, with what it found. When the reply is done, the steps fold into one line.
• Searches that match nobody: A search that matches nobody stays in the conversation, which asks which part to widen. A button opens it anyway.
• Other searches: A request for funding rounds, hiring, a technology, local businesses, similar companies or subdomains gets a button that opens that search empty.
• One conversation across pages: The AI Assistant panel beside the search opens on the same conversation, so you can keep refining there.
• Conversations on your account: Conversations are saved to your account, private to you, and follow you to any device. Recent panel conversations kept in your browser move there once.

​

Find leads

• Suggestions under the box: Three ideas under the search box open their search straight away with the filters set. With business context written, they are tailored to it. See Business context.
• A taller search box: The box grows as you type, Enter sends, and Shift+Enter adds a line.
• No filtering as you type: Typing leaves the page as it is, and every request goes to the AI Assistant, including the name of a search.
• Suggested for you retired: The suggestions under the box replace the Suggested for you cards.

​

Find leads

• Row prices up front: Company activity and Similar companies searches charge for the rows they add, and the price per row shows before you run one. See People and companies.
• What a search adds: Those searches list every column the new table will carry, grouped and named, in place of a preview of rows.
• One cost line: The enrich step counts the rows and the enrichments you picked in one figure, so nothing reads as free that is not. See Choose what to enrich.

​

Senders and mailboxes

• Daily send limit on the page: Change a mailbox’s daily send limit from its own page, and read the date the new number starts applying before you save. See Mailbox health and capacity.
• Warm-up you can scan: The mailboxes list gains a Warm-up column, and the mailbox page a matching card, each showing how far the ramp has to go.
• Warm-up after a raise: Raising the limit on a mailbox that already finished its ramp no longer puts it back on one.
• Three status filters: Filter the list by verified, connected or disconnected, the same three states the Status column shows.
• Connected via: The list says how each mailbox was connected, Direct or InboxKit, instead of naming the same mail service on every row.
• Bounces on the chart: A mailbox’s activity chart plots sends, real replies and bounces, and a tooltip says what each one counts.
• A warning you can dismiss: The banner counting disconnected mailboxes can be closed, and returns on its own when another mailbox drops.

​

LinkedIn

• Withdraw a connection request: A step that withdraws an invitation still pending now runs in the browser, pressing the same controls a person would. See LinkedIn steps.
• More steps to try by hand: From a sender’s LinkedIn tab you can queue a message or a withdrawal, beside the profile view and connection request already there.
• Its own tab group: The tab Oneprofile works in sits in a named tab group, apart from your own LinkedIn tabs. Chrome asks before it may do this. See Connect LinkedIn.
• Connection token in view: The connection token shows on the sender’s LinkedIn tab without a click, ready to copy into the extension.

​

Agents

• Rewritten message agents: The built-in email, LinkedIn and multi-channel agents make the whole case in the first message, then keep each follow-up shorter. See Built-in agents.
• Problems in their words: The first email names two or three problems drawn from the person’s role and company, rather than listing features.
• Messages that read on a phone: LinkedIn messages carry no bullet lists and no line breaks.

​

Bug fixes

• LinkedIn steps after a redesign: Steps run again on LinkedIn’s rebuilt profile and feed pages, which had stopped every step that opened one.
• Step details: A finished LinkedIn step explains itself in a sentence, instead of showing an internal code.
• Support chat: The chat launcher stays on screen after you leave Tables, Integrations or Billing.
• Dates without a time: A date that carries no time reads as the day it is, instead of the day before.
• Mailboxes on a sender: Clicking a mailbox row on a sender opens that mailbox’s page, and the back arrow returns to the tab you came from.

​

Find leads

• Five searches up front: The Find leads page opens on Find people, Find companies, Similar companies, Local businesses and Companies by technology. See People and companies.
• Company activity: Twenty featured signals sit on the Find leads page, with every signal one click away. Each card opens a search on that one signal.
• One signal, one subtype: A signal search runs on the signal you picked and, where the signal offers them, one subtype. The subtypes are listed in the open, and the search waits until you pick one.
• Signal columns: A table built from a signal search carries the signal’s own details as columns, such as a post’s text and link, a funding amount, or a filing date, beside the standard name, title, company and domain columns.
• Content filters where they apply: Keyword filters on the text behind a signal appear only for signals that carry text, and only once the subtype is chosen.
• A preview that keeps the rest hidden: Preview 5 rows shows each row’s name, job title and company. Every other column is filled in when the rows are added to a table, and never leaves the server before then.
• Review profiles retired: The review profile search is gone. Company searches and the Company activity signals cover what it found.

​

Senders and mailboxes

• Teammate mailbox links: Generate a connect link from any sender profile and send it to a teammate. They sign in with Google on a page of their own, and the mailbox is added to your workspace. See Connect a mailbox.
• Link expiry and revocation: Links expire after a week, and you can revoke every link for a sender at once.
• Sending domain reporting: Export every mailbox in the workspace as CSV, or the same numbers rolled up per sending domain. See Mailbox health and capacity.
• Date ranges per mailbox: Set a date range and the chart, recipient domains, and bounces all follow it.

​

Sequences

• Sending schedules: New sequences start on weekday working hours in your timezone. See Schedule and sending window.
• Timezone as its own choice: The timezone is separate from the schedule preset, so changing one keeps the other.
• Named rosters: Sequence rosters lead with a name and company instead of a bare email address. See Choose the audience.
• Funnel analytics: Sequence analytics read as a funnel, from everyone enrolled through to revenue. See Sequence analytics.

​

Inbox

• Goal-directed AI writing: Pick what a message is for, or write your own goal, and the AI writes toward it. Attach a call-to-action link and the label it should read as. See Reply with AI.
• Replies in the contact’s language: Have a reply written in the language the contact used.
• Kept drafts: Unsent drafts are kept per conversation and restored as written. See Inbox.
• Labels and channels: Labels take an icon and a color as you create them, and every message shows its channel.

​

Enrichment

• Manager lookup: Add a Manager column with an email address or a LinkedIn URL, and get the person one step up. It returns the manager’s name, title, LinkedIn profile, and work email.
• Cost before the run: The cost shows in the column picker before you run it. See What enrichments cost.

​

Bug fixes

• Cell run details: Clicking a cell opens the run behind it whenever there is one, including cells that came back empty.
• Column groups: Opening a cell on the right of a wide table no longer slides the grid sideways when a group is collapsed.
• Editing a cell: Opening an editor on a cell near the edge of a wide table keeps the table where it was.
• Cell errors: A failed cell’s tooltip reads in plain English instead of showing the internal error name.
• AI columns: A column you build yourself shows its answer in the grid instead of a folded group with every value hidden.
• Confidence marks: The grid marks only the answers an agent hedged on, and no longer vouches for cells the agent said nothing about.
• Agent inputs: A required input binds to the column that actually holds the fact, so a person agent no longer researches the company instead.
• Agent activity: Re-running a cell shows the new run’s steps, not the previous run’s.
• Adding an agent to a table: The dialog keeps the mappings it suggests, and the ones you type.
• Tables list: Coming back to Tables shows the real row counts, instead of a snapshot taken before the rows were added.

​

Find leads

• People and company search: Filter people by title, past roles, location, skills, school, and time in role. Filter companies by industry, headcount, revenue, keywords, and headcount growth. See People and companies.
• Match count and preview: See the match count and a preview before you spend anything.
• Saved searches: Save a search, put it on a cadence, or refine it in plain language. See Saved searches.
• CSV upload: Preview a CSV and the table it would create before you commit. See Import a file.
• Records from a connected tool: Sample real records before adding them. Keep the table attached to its source and refresh it on a schedule. See Import from a tool.
• Subdomains as rows: Hand Oneprofile a domain and get its subdomains back as a list. See Subdomains.

​

Tables

• Saved views: Save named views per table with their own filters, sorts, and hidden columns. See Views.
• Tags, stars, and recents: Tag tables, star the ones you use, and jump back to recent ones. See Create a table.
• Column templates: Capture a set of columns as a template and apply it to another table. See Save and reuse a column.
• Re-run failed cells: Re-run every failed cell in a table in one go. See Run columns.
• CSV export: Export any table as CSV.
• Trash: Move a table to the trash and restore it later. See Trash.

​

Enrichment

• The catalog: Enrichments cover emails, phone numbers, LinkedIn profiles, and company and person records. See What you can enrich.
• Cost in the column picker: Every enrichment shows its cost before you run it. See What enrichments cost.
• Cell receipts: Open any cell to see where the answer came from and what it returned. See Where the data comes from.

​

Agents

• Built-in agents: Research a row for what a company does, who they sell to, whether they are hiring, and ICP fit. See Built-in agents.
• Drafted from a description: Describe a task and watch Oneprofile draft the agent for it. See Build an agent.
• Versions: Save an agent, version it, and revert to any earlier version. See Agent versions.
• Your own MCP servers: Add your own MCP servers so an agent can reach systems only you have. See MCP servers.
• Training rounds: Train an agent by critiquing real rows, then approve the revised prompt. See Train an agent.

​

Outreach

• Sequences from a table: Point a sequence at a table or a view and Oneprofile enrolls those rows. See Choose the audience.
• Email and LinkedIn steps: Steps send an email, view a LinkedIn profile, send a connection request, or send a LinkedIn message. See Build a sequence.
• Branches and waits: Branch on replies, bounces, missing emails, and accepted invitations, with waits between steps. See Sequence conditions.
• Drafted cadences: Describe the cadence you want and get a full draft back, then edit any step.
• Step previews: Preview a step rendered for a real contact, or send it to yourself as a test. See Email steps.
• Pre-flight checks: Catch missing senders and empty steps before anything sends. See Launch checks.
• Per-step reporting: Track opens, replies, and bounces per step, and export it all as CSV. See Sequence analytics.
• LinkedIn in your own browser: A Chrome extension runs LinkedIn steps in your own logged-in session. See Connect LinkedIn.
• Limits and pacing: Set daily limits and pacing per sender, and watch the queue of upcoming steps. See LinkedIn steps.
• One pause switch: Pause all LinkedIn activity from the app or the extension.

​

Inbox

• Replies in one place: Replies to your sequences arrive in Oneprofile instead of scattering across mailboxes. See Inbox.
• Mailbox sources: Connect Google Workspace mailboxes, or add the ones you already run in InboxKit. See Connect a mailbox.
• Labels and AI drafts: Label threads, reply with an AI draft you approve, and mark threads read in bulk. See Reply with AI.
• Slack and email alerts: Get a message when a reply arrives or a mailbox disconnects. See Notifications.

​

Integrations

• Write rows back: Map columns to a destination record type and write them into a connected tool. See Export to a tool.
• Dry runs: Dry-run an export against real rows and see exactly what would be written.

​

Workspace and account

• Spend breakdown: Break spend down by table, column, run, or agent, and follow your balance over time. See Usage and reports.
• The raw ledger: Read the ledger entry behind any charge, and export a usage report as CSV.
• Credit packs and invoices: Buy credits when you need them, and see every invoice. See Credits and billing.
• Referrals: Share your referral link and earn credits when someone you invite joins. See Referrals.

​

Bug fixes

• Long answers: A cell holding several paragraphs saves, instead of failing every other cell computed with it.
• Table grid: The grid stays put instead of blanking to a loading state and jumping back to the top every few minutes.
• Run status: A table whose last run failed or was cancelled says so, instead of going quiet as though nothing had happened.
• Pasting into a table: A paste that runs past the last row adds the rows it needs, instead of dropping the extra values.
• Large pastes: A paste big enough to fill thousands of cells writes in one go instead of erroring.
• Formula editor: Changing a step’s function keeps the settings the new function also uses.
• Long runs: A slow column keeps going past the first few dozen rows, instead of timing out and discarding its work.
• Row actions: Every row shows its run button, disabled with a reason on a table that has nothing to run.

​

Tables

• Typed columns: Add text, numbers, true/false, dates, timestamps, lists, and nested objects. See Field columns.
• Automatic recalculation: Edit a cell and every formula depending on it recalculates on its own. See Reference other columns.

​

Columns

• Chained formulas: Chain several steps in one formula, across text, numbers, dates, lists, and if/then logic. See Function columns.
• Preview against real rows: Preview a formula against your real rows before you save the column.
• Fallbacks: Choose what a cell falls back to when a step fails: the raw value, empty, or a value you set.

​

Runs

• Run all, retry failed, or spot check: Re-run a column over every row, only the rows that errored, or the first ten. See Run columns.
• Progress and cancel: Watch recomputes in a progress banner and cancel a run you started by mistake.