Nº 01 · Changelog, page 3
Earlier changes, page 3.
Changes shipped from 19 Oct 2026 to 20 Oct 2026, newest first.
Applications open soon
See the whole workspaceEarlier changes
October 2026
Activity pages recover from unsupported date ranges
Admin Activity pages now use the normal 30-day window when a URL contains an unsupported range, including inherited object names. The 30-day, 90-day and 12-month choices are unchanged.
Ask the Agency Agent which skills you have pushed to clients
The Agency Agent and the organisation-wide connector can now list the agent skills an agency has pushed, with each client's copy shown as locked or unlocked and whether it has the latest version. Owners and admins only.
Agency Skills page lists the ready-made actions
The agency Skills page now shows the built-in actions as locked cards. Copy one as a skill, adjust it, then push it to your clients.
Connected-app steps open with their fields, logo and action name
A connected-app step in an automation now shows the app's logo and name as soon as the automation opens, on the canvas and in the step panel. The chosen action reads as its name and description with a Change button instead of a slug and a long list, and its arguments appear as labelled fields for every action, not only the first forty in the list. Saved steps have their fields fetched as the automation opens, so opening one shows its fields straight away; if they are still on their way you see a placeholder rather than a JSON box.
Safer captured-answer choices in automations
The data picker no longer offers misleading individual tokens for captured keys containing dots or token syntax. It keeps their complete containing object available and explains the limitation, so a literal dotted answer cannot be confused with a different nested answer. Existing saved tokens are unchanged.
Internal notes on a booking page
Write notes for your team on a booking page, such as who to loop in or what to check first. They show on the page's editor and at the top of every call booked through it. Visitors never see them: not on the booking page, in emails or in the calendar invite.
Retry calendar and import tools without losing your work
Copy-hours controls show a loading state and a local retry when their editor cannot load. Import mapping has a retry that keeps your selected file and mapping. CSV parsing starts after file selection, reports read errors in the upload area, and prevents a delayed previous selection from replacing the latest file.
Name the calendar event for a booked call
Each booking page can set the title your own calendar shows for a booked call, such as the page, the lead's name, their email and the length, with a preview as you type. Left blank, the title stays as it was. The attendee's invite keeps the page name.
Claimed-source revenue counts every payment in a long range
On a busy workspace, the claimed-source attribution view totalled only the first thousand payments in the chosen range and showed that as the revenue. It now reads every payment before splitting claimed from unclaimed.
A record added in a filtered Data view keeps the view's values
In a Data table, Add record inside a view that filters a select field to one option, a multi-select to chosen tags, or a checkbox to ticked now starts with those values, so the new row stays in the view instead of disappearing. A note under the toolbar says which fields were filled. Filters on numbers, dates and links, and fields two filters disagree about, are left empty.
Strengthened framework and dependency safety checks
Framework and supporting packages move to reviewed security releases. Installed-consumer tests cover image conversion, CSS compilation, address classification and routing. A local nesting-depth guard protects the glob parsers used by lint and styling tools; its upstream advisory remains tracked rather than treated as cleared. Lead profiles load the Notes editor only when its tab opens, with loading and retry states.
Events follow your account timezone automatically
A new event uses your account timezone and keeps following it: change the timezone in Settings and every event that follows it moves too. Pick a specific timezone on an event and it stays on that one until you choose Account timezone again. If your account has no timezone yet, the event editor offers to set it from your device in one click.
A new form always gets a free link
Creating, copying or importing a form, or renaming its link, now checks the link against every workspace before saving. A name like Application no longer fails with a database error when another workspace already holds that link in a draft.
Tag a teammate in a lead's notes and resolve the thread
A lead's Team notes are now a conversation. Reply to a note, type @ to mention a teammate, who is told in the app and by email, and resolve a thread once it is dealt with. Handy for handing a lead from setter to closer. Comments also show in the lead's timeline.
Edit or remove a comment on a lead, and reach it from a deal
You can now edit your own comment on a lead, marked as edited, and remove one nobody has replied to. Owners and admins can remove any comment. A deal's card menu and side panel link straight to its lead's notes, where you can mention a teammate.
Lead profiles load their history alongside other details
Opening a lead starts calls, tasks, money and other profile reads without waiting for its journey history first. The drawer and full page still show the same complete profile, and failed history reads remain visible errors rather than empty journeys. Authorised drawer responses include private numeric server timings to help diagnose remaining delays.
Automation failures withhold provider and parser details
Failed Slack, Telegram, single-recipient email and Connected App transport requests use fixed errors without copying provider exceptions or causes. Connected App lookup and malformed JSON errors also withhold upstream and parser text. Agent reference search and resolution logs likewise withhold raw database and getter exceptions while retaining successful references; thread rename and archive logs withhold returned database messages. The outer Agent stream diagnostic withholds raw exceptions and causes without dropping earlier events or changing its safe failure response. Approved-action result-save logs also withhold exception details while retaining the check-before-retry warning and refusal to rerun the action. Optional email-history records withhold provider failure text, and their outer record-save logs withhold exceptions without changing delivery results. Scheduled-report error records and audit reasons use fixed diagnostics; the report panel hides unfamiliar older errors without rewriting rows. Scheduled-report claim and outcome-save logs also withhold returned database errors, retaining the report id and existing send behaviour. Ownership checks still block unverified accounts. Successful payloads, HTTP diagnostics, timeouts and retry policies stay unchanged. Uncertain delivery is stated without claiming the app never acted; email booking links remain valid and segment broadcasts retain safe counts.
Typeform, Cal.com and Fathom cards say what they bring in
These three apps have a SIGNL card and a Composio card on Integrations. The SIGNL card now carries a label for what flows in (responses, bookings, transcripts), beside the Composio card's "Automation actions via Composio", so the two read as a pair.
Pulse shows retention by cohort
The Admin Portal Pulse page has a retention grid: for each month organisations first paid, how much of that first MRR is still paid at the end of each later month, per currency. Organisations already paying when counting began form one opening cohort. It fills in as months of daily snapshots build up.
The setup chat shows Whop as connected when you signed in with Whop
If you connected Whop with Sign in with Whop, the setup chat's Whop card now says Connected, the same as the Integrations page, instead of Not connected. The public integrations docs also now point to Integrations, Payments for processor webhooks.
Add or remove ways to sign in to your account
Settings, Account now lists how you can sign in (password, Google, Whop). Set a password on an account you made with Google or Whop, connect Google or Whop to a password account, or remove a method you no longer use. The last way in can never be removed, each change is recorded in the audit log and emailed to your account address, and a two-factor code is asked for first if you use one.
Ad platforms get the real time of a sale
When a payment is sent to Meta, Google Ads or TikTok, it now carries the time the sale happened, including a send that is retried later. Before, a retry was stamped with the time of the retry. Meta refuses events older than seven days, so a retry that late is skipped and the send log says why.
Add record shows that it is working
In a data table, the Add record button now reads Adding with a spinner while it saves, so a slow save no longer looks like nothing happened.
Leaderboard adjustments and webhook replays are recorded in the audit log
Adding or removing a manual leaderboard adjustment, and replaying a webhook delivery, now each add an audit entry with who did it and when. Webhook payloads are never copied into it.
Censor mode hides lead names on the agency calendar
The agency calendar, which lists every client's calls in one place, showed each lead's name beside the time even with Censor on. The name is now blurred like the rest of the page, so a screen share of it no longer reads out who is booked across your clients.
Push Agent skills from your agency to every client
Write a skill once under Agent skills in the agency view and push it to the clients that need it, locked or unlocked. Clients with an unlocked copy follow your updates until they edit it; a locked copy can only be changed by you.
Choose which AI labs the Agent may use
Account owners can tick the AI labs the Agent is allowed to use, in Settings, Organisation. The model picker then lists only models from those labs, in every workspace.
Ask the Agent about the apps you have connected
The Agent can now see which apps your workspace has connected and read from them, for example the rows of a Google Sheet or a list of Gmail drafts, then answer from what it finds. It is read-only for now: it only runs tools that read, and changing something in an app is not available yet. Connect an app on the Integrations page first.
Choose the model the Agent starts on
Account owners can pick the default model for the Agent in Settings, Organisation. Every workspace starts on it, and anyone can still choose another model from the picker. If the model is later withdrawn, the Agent falls back to the standard default.
The Agent counts every lead, call and payment in a large workspace
Ask the Agent about a workspace with more than a thousand leads, calls, deals or payments and its totals, hot lead count, average intent and 30 day figures now cover every record, not the first thousand. If part of the data cannot be read, the Agent says so instead of quoting a partial total.
Mention tasks, automations, offers and pipelines in the Agent
Type @ in the Agent and pick a task, an automation, an offer or a pipeline, as you already can a lead, call or deal. The Agent reads that record for the answer. You only see the types you have permission to read, and a record from another workspace never resolves.
The Agent's model picker shows which models are Verified
Each model in the Agent's picker now says whether SIGNL has checked it against the Agent's tools. Verified models passed: they call tools and answer in words. Unverified ones have not been checked yet and may come back empty. A model that keeps failing is taken out of the picker automatically.
Save the prompts you reuse as Agent skills
Admins can write a skill once, pick the tools it may use and run it from the / menu in the Agent. Each save keeps the earlier version, and a teammate only sees a skill if they are allowed to use every tool it lists.
Ask the Agent to look into several things at once
The Agent can now hand a research task to a helper that reads your data on its own and reports back a short summary. In the Agency Agent, each helper works inside one client, so a question across your clients runs side by side. Helpers can only read: if one finds a change worth making, the Agent proposes it and you approve it as usual.
Keep the Agent to models that keep none of your data
Account owners can require zero data retention for the Agent. It is on by default, and the model picker then offers only models that a provider serves without keeping your prompts and replies. Turn it off in Settings, Organisation, to see every model.
Pick your Airtable base and table from a list
Airtable steps now let you choose the base and the table from the ones your connected Airtable account can see, instead of pasting an ID. If the list can't load, you can still type the value.
The automation editor's buttons are reachable on a phone
On a phone the automation editor's top bar ran off the screen: the back link overlapped the List button, and Runs, Versions, Test run, Discard and Publish could not be reached. The bar now wraps onto more than one row.
A Runs table for every automation, with a strip of what each step did
Open the full table from an automation's Run History. Each of the last 50 runs shows one coloured cell per step and a line saying what happened, such as Failed at Send Email. Choose a run to see each step's result beside the table.
The empty-drafts cleanup now finds automations you emptied again
An automation where you added a step and then deleted it is saved as an empty draft, which the Automations list used to treat as someone's unfinished start. It now counts as empty, so the empty-drafts banner offers to delete it. Anything with a step in its draft, or that has ever run, is still left alone.
The Automations list no longer scrolls sideways on a phone
On a phone the Automations list made the whole page wider than the screen, so the page slid sideways. Only the table inside its card scrolls now, and the rest of the page stays put.
New booking pages use the hours you see in the editor
A new booking page showed Monday to Friday, 9 to 5 in its editor, but until you saved it visitors could book any day, including weekends. New pages now start with those hours saved, so what the editor shows is what visitors are offered.
Booking pages without a saved time zone offer the hours the editor shows
A booking page that had hours but no time zone saved showed 9 to 5 in its editor, yet visitors were offered 10 to 5:30 in UK summer time, because the slots were worked out in UTC. They now use the same default zone as the editor, Europe/London, so the page offers exactly 9 to 5.
Hide the logo strip and headline on a booking page
Each booking page now has two switches under Branding: show the logo strip, and show the page name as the headline. New booking pages start with the logo strip off, like public forms. Pages you already have look exactly as before until you change them.
Choose several booking pages for a booking automation, right on the canvas
A booking automation set to run on only the booking pages you choose now shows a tick box for each page by name, so you can pick several without leaving the automation. Draft pages are marked, and a deleted page is shown so you can untick it.
A guide and a connector file for building your own dashboard on SIGNL
A new guide shows how a dashboard you built yourself reads SIGNL with one read key: the snapshot, the leads with their first touch, and your content items with clicks, opt-ins, calls booked, revenue, shows, wins and cash. It comes with a connector file you can drop into your project that never throws on an outage.
Censor mode now hides lead names on the calendar grid
With Censor on, the month grid on the Calendar page still showed each lead's name beside the time, and the day view's Overlaps with line named the other lead. Both are now blurred along with the rest of the page's names, so a screen share of the calendar no longer reads out who is booked.
Pick any timezone for a calendar and copy hours to chosen days
A calendar's timezone is now a full list with GMT offsets, your account timezone is marked and one click puts it back, and the hours header says whether a calendar uses your account timezone or its own. The copy button beside a day now opens a list so you can copy those hours to all days or just the ones you tick.
Route an automation on a call's predicted grade
A Condition or Paths step can now check a call's predicted grade (high, medium or low) and score, for events that follow a booking such as a no-show or a cancellation. It is a check only: nothing acts on a grade by itself.
Setup checklist asks whether your call reminders are on
The onboarding checklist has a seventh item, Call reminders switched on. It is done once an Event Flow is active, so booked calls get their reminder messages, and it links to Event Flows. A paused flow does not count.
Choose how you get your two-factor code
Turn on two-factor sign-in with an authenticator app, a text message to your phone, or both. At sign-in you pick whichever method you have to hand, and you can add or remove a method from your account settings at any time.