> For the complete documentation index, see [llms.txt](https://www.cakewalk.security/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://www.cakewalk.security/docs/changelog/readme.md).
# 2026
{% updates format="full" %}
{% update date="2026-08-26" tags="new-releases,improvements,fixes" %}
## Google sign in for six more apps
Six more apps can be provisioned with a [Google login](https://www.cakewalk.security/docs/human-access/how-to-guides/auto-provisioning/how-it-works/sign-in-with-google) instead of a stored password. Sentry, Figma, Dovetail, Zoom, Loom and Slack join ChatGPT, Miro and Mixpanel.
Sign in with Google is enabled per app, so coverage grows as each one is proven. Existing configurations keep working untouched.
## Group membership with a choice of action
* Adding users to a [User Group](https://www.cakewalk.security/docs/human-access/how-to-guides/users-and-groups/groups-and-role-based-access-control-rbac) now offers three paths: raise access requests for each person, apply the access directly as an Admin, or record the membership and change no access at all.
* The same choice applies when removing members, so a group change no longer forces one behaviour on everyone in it.
* Where an option is unavailable the dialog says why, for example while the group is still onboarding or when the group has no access to the app in your identity provider.
## Broken integration warning in App Governance
* A warning icon in [App Governance](https://www.cakewalk.security/docs/human-access/how-to-guides/apps/app-governance) marks any app whose integration has broken, so a silent sync failure is visible from the overview rather than discovered later.
* Filter the table by Yes, No or No Integration to see only the apps that need attention.
* App detail carries the same warning, plus a separate notice when the last sync ran too long ago.
## Access Sync coverage
* [Access Sync](https://www.cakewalk.security/docs/human-access/how-to-guides/access-sync) reads users and their access from [Vercel](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/vercel), [incident.io](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/incident.io), [Sanity](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/sanity) and [PagerDuty](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/pagerduty).
## Agent Cake coverage
* New apps: [Sanity](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/sanity), [ClickHouse](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/clickhouse), [Heap](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/heap) and [PagerDuty](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/pagerduty).
* [Sentry](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/sentry) and [Figma](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/figma) rebuilt on browser automation, which is what allows them to sign in with Google.
## A better Slack app
* Agent Cake answers in Slack on an upgraded model.
* Permission descriptions render properly in request threads, and app owners appear by name rather than as a raw email address.
Improvements
* Bulk task claiming keeps the claims that succeed rather than failing the whole batch, and tells you which tasks did not go through.
* An auto-provisioning configuration only needs an inbox when the app signs in with an email address and password. Configurations that sign in with Google no longer ask for one.
* User sync from your HRIS can be filtered by email domain, so a shared directory only brings in the people you govern.
* Bulk onboarding handles longer lists without timing out.
* Offboarding completes more reliably, and onboarding and offboarding request logs are paginated and filterable by date.
* Permission levels can be edited on Agent Cake provisioned apps inside identity provider user groups.
* ClickUp keeps its session between provisioning runs, which makes runs faster and less likely to trip a login check.
* Curated Connection Access, which lets an Admin choose which Connections each User Group may use, has its foundations in place.
Fixes
* ChatGPT no longer fails on every call after connecting. Its MCP client sent a protocol version the gateway rejected.
* Atlassian sign in failed when Atlassian served a security key as the second step.
* Calendly reported a generic failure when the account had no free seats. It now says so.
* Vercel permission levels were wrong in the setup form.
* Loom sign in now completes when two step verification uses an authenticator app.
* Invitations could be sent twice to the same person.
* User sync could double apply an update from your HRIS, and a race between two syncs could leave the wrong result.
* An app template could be mapped to the wrong app.
* Slack provisioning failed on an input that had been mislabelled.
{% endupdate %}
{% update date="2026-07-31" tags="new-releases,improvements,fixes" %}
## Sign in with Google
Some apps only let people in through Google. Agent Cake can now [sign in with a Google account](https://www.cakewalk.security/docs/human-access/how-to-guides/auto-provisioning/how-it-works/sign-in-with-google) instead of an app username and password, which brings those apps into Auto-Provisioning for the first time.
Connect a Google account once in Settings and reuse it everywhere it applies, or enter Google credentials on an individual app's own configuration screen. Live for ChatGPT, Miro and Mixpanel.
* A configuration can offer a username and password login and a Google login at the same time, rather than forcing a choice between them.
* This drives Google's own sign in page rather than an OAuth consent screen. The account has to allow password sign in, and where 2 Step Verification is on it has to use an authenticator app.
* The [Set up Service Accounts](https://www.cakewalk.security/docs/human-access/how-to-guides/auto-provisioning/how-it-works/set-up-service-accounts) guidance still applies. Use a dedicated account, not a person's own login.
## Company SSO login for every customer
* Logging in to Cakewalk through your company's own SAML identity provider is available to all customers. It was previously limited to some.
* Signed authentication requests are supported for companies that require them.
## Public API additions
* Read only API credentials, for integrations that should never write.
* [Webhook signatures version 2](https://www.cakewalk.security/docs/human-access/open-api-and-mcp/webhooks/webhooks-signature-verification), which adds replay protection.
## ChatGPT
* Agents running in ChatGPT can use Cakewalk. Three tools stand in for your whole catalog, which is how a client without native MCP support reaches your company's apps.
## Human Access and Agent Access together
* Users and User Groups stay in sync across both modules, so a person added in one appears in the other.
* Navigation moves straight between the two, and a module's entry appears only if your company uses it.
## When an app has no seats left
* Provisioning that fails because the target app has run out of seats now says exactly that, rather than reporting a generic failure, and the requester is notified.
Improvements
* New Agent Cake apps: [Twilio](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/twilio), [Lemlist](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/lemlist), [Sigma](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/sigma) and [Momentum](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/momentum). Twilio arrived as a full provisioning suite.
* Agent Cake handles a one time code from an authenticator app during sign in.
* OpenAI Platform provisioning moved to the vendor API.
* A Connection whose refresh token has expired shows a reconnect button instead of failing quietly.
* An audit event with a large payload can be downloaded in full.
* Onboarding no longer times out on companies with long user lists.
* Deactivating a user revokes their agent and Connection access in the same step.
* Approval prompts in Claude Code accept correctly.
Fixes
* Notion deprovisioning reported people as removed when they were not.
* Miro deprovisioning failed for users with no team.
* Zoom created a duplicate user in some configurations.
* Metabase test connection failed on valid credentials.
* Amplitude sign in broke after the vendor changed its login page.
* Bulk approval of access requests could race and drop a request.
* Adding a user who had already been deactivated raised an error.
* Offboarding on a slow app timed out before finishing.
{% endupdate %}
{% update date="2026-06-28" tags="new-releases,improvements,fixes" %}
## Add your own Connections
The catalog is no longer the limit. Admins can add [custom MCP servers](https://www.cakewalk.security/docs/ai-agent-access/how-to-guides/connections/add-a-custom-connection) that Cakewalk does not ship, and they behave like everything else: same Policies, same approvals, same Audit Log.
Each custom Connection carries a name, a category and an image, so it sits in the catalog rather than beside it.
## Connections that need setup are labelled
* A Connection that cannot register itself automatically is marked as awaiting setup, with a warning icon in All Connections and an alert card explaining what is missing.
* My Connections disables Connect on anything not yet configured, so an employee does not walk into a dead end.
* Mark as configured moves the Connection to Active once the credentials are in place.
* Search excludes Connections that are still awaiting setup.
## Correct a tool's action type
* Admins can [reclassify what a tool does](https://www.cakewalk.security/docs/ai-agent-access/concepts/policies), as Read, Write, Destructive or External, one at a time or in bulk from the Connection's Tools tab.
* Policies are keyed to action type, so this is the fix when the wrong Policy fires because a tool arrived classified incorrectly.
* The correction survives the next catalog sync.
## Personal access tokens
* A Connection can authenticate with a personal access token. That completes three ways to connect, alongside automatic client registration and credentials an Admin supplies.
## Public API for access requests and the audit log
* Create access requests through [the API](https://www.cakewalk.security/docs/human-access/open-api-and-mcp/api-reference/requests), to grant, change or remove access, routed through Policy and approval exactly as they are in the product.
* Read requests and audit events through a filterable endpoint, which gives a SIEM something to pull from on its own schedule.
## Denials that explain themselves
* When a Policy denies a tool call, the response names the Policy that fired.
* An agent client that cannot show an approval prompt gets a message saying so, rather than an unexplained denial.
Improvements
* Employee onboarding for Agent Access runs end to end for companies with the flow enabled.
* User sync from your HRIS runs continuously and sends only what changed.
* Policies live in one tab at settings level, with sub tabs for custom and default Policies.
* Agent aliases resolve names the search misses, for example Open AI MCP to ChatGPT.
* Agent and Connection tabs renamed so My and All are unambiguous.
* Add Connection and Add Agent state that only company sanctioned items are listed.
* Agent Cake gives better answers, from reworked templates.
* [Lovable](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/lovable) and [Miro](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/miro) rebuilt, Lovable on a new automation engine.
* Collapsible tables keep the first column in place while you scroll sideways.
Fixes
* Agent sessions survive a gateway restart rather than being lost.
* User import failed with a server error when the person already belonged to groups.
* Users who had been ignored could not be onboarded later.
* Far future termination dates from Personio were read as immediate.
* Invitation login failed for some users.
* Two Connections can share the same URL, which previously raised a conflict.
* Reactivating a Connection returned a server error.
* Excessive numbers of deactivated users were pulled in from Datadog.
* A cron trigger for Intercom stopped firing.
* Task assignment emails named the wrong person.
{% endupdate %}
{% update date="2026-05-28" tags="new-releases,improvements,fixes" %}
## Connect once, inherit your access
Agent Access is live. Employees authorize an app once, and every Agent they run reaches it through Cakewalk with exactly their own access, governed on every call.
The Agent never holds the credential. Cakewalk injects it at call time and never returns it to the Agent or the model, so a compromised context has nothing to steal.
* Agent permissions match that person's permissions in each app, so a teammate's Agents stay scoped to the teammate.
## Inline approvals
* A sensitive action [pauses inside the Agent client](https://www.cakewalk.security/docs/ai-agent-access/concepts/the-tool-call-lifecycle). Approve or deny in place and the Agent resumes with full context.
* The Agent experiences a slow tool call, nothing more, so a decision does not end the task.
## Policies
* Create custom [Policies](https://www.cakewalk.security/docs/ai-agent-access/concepts/policies) and review the defaults, each evaluated per company.
* Conditions read the person's department, job title, team and location, and the app's own attributes.
* Test a Policy against past agent activity before it goes live, and see which calls it would have decided differently.
## Sessions and the Audit Log
* Every tool call through the gateway is [recorded](https://www.cakewalk.security/docs/ai-agent-access/concepts/audit-log): who ran it, what it did, against which Connection, and which Policies fired.
* Sessions are viewable across Agents, users and Connections. Admins see everything, employees see their own.
* Collapsible rows show the tool calls inside a session, and a side panel opens one call in full.
## Agents and Connections
* Agents overview and detail, in grid and table views. My Agents for employees, All Agents for Admins.
* Connection detail lists the tools that Connection exposes, with disconnect and bulk actions.
* [Agent status](https://www.cakewalk.security/docs/ai-agent-access/concepts/connection-and-agent-statuses) derives from real activity: Active, Inactive or Not Connected.
* Add an Agent before it ever connects, so the catalog reflects what your company sanctions rather than only what has been used.
Improvements
* An Admin can supply OAuth client credentials for an app that cannot register a client automatically.
* Access Reviews no longer time out on large runs.
* Bulk approvals are batched, which stops large approvals failing part way.
* Apps and Agents can be discovered and imported from a Google account.
* Agents that are not connected can be removed.
* Sessions tables sort newest first by default.
* Company properties can be managed from settings.
Fixes
* Connect is disabled on a Connection whose credentials are missing, rather than failing at sign in.
* The first setup step shown for Claude was wrong.
* Onboarding failed with an invalid task error.
* Groups from your HRIS were not created when users were imported.
* A duplicate permission level raised a conflict on save.
* Copy to clipboard triggered the row action behind it.
* Figma deprovisioning failed on some accounts.
* Amplitude picked the wrong project when several were available.
* Calendly showed its terms banner too briefly to accept.
{% endupdate %}
{% update date="2026-04-30" tags="new-releases,improvements,fixes" %}
## Empty states redesigned
Every table now explains what is missing and what to do about it, instead of showing an empty rectangle. A quieter month on Human Access while Agent Access was being built.
* The redesign covers users, apps, requests, tasks, reviews and group imports.
## Clearer ownership on onboarding and offboarding
* The onboarding and offboarding tables label the column Assignee(s) rather than Manager, which matches who actually does the work. A manager is often not the person completing the task.
Improvements
* User import reports duplicates and a count before anything is created.
* Notification centre filters work correctly for Slack broadcasts.
* A failed app integration shows a reusable alert with the reason, rather than a generic error.
* [Calendly](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/calendly) rebuilt on browser automation.
* Slack user data is cached with an eviction policy, which makes the Slack app noticeably faster.
* Leaving a page cancels its in flight requests, so stale results no longer arrive after you navigate away.
Fixes
* Task dates broke when the browser translated the page.
* Access Sync stopped running for Linear on some accounts.
* Agent Cake sign in fixes for Lovable, Notion, Loom and Atlassian.
* Server errors on Access Reviews and on the audit endpoint.
* The Slack app install link from the website was broken.
* User sync through an app integration failed for some accounts.
* Importing apps without an owner failed.
* A timeout and a dropped connection were treated as the same error, so retries behaved wrongly.
{% endupdate %}
{% update date="2026-03-27" tags="new-releases,improvements,fixes" %}
## A rebuilt Slack app
The Slack app is rebuilt around conversation. Cakewalk runs as an [assistant in your Slack sidebar](https://www.cakewalk.security/docs/human-access/connections-and-integrations/cakewalk-for-slack), understands plain language, and keeps everything about a request in one thread.
Most access work no longer needs the web app at all.
* Ask for what you need. Request access for yourself or for someone else, change a permission level, remove access, or ask for several apps at once.
* Check the status of a request, have a reminder sent to your approver, and find out who owns an app.
* When the app you asked for is not tracked, the assistant suggests alternatives rather than stopping at a dead end.
* Approvers complete tasks in Slack. Every follow up stays in the same thread, and live status shows what the assistant is doing while it works.
* Mention Cakewalk in a channel and it moves the conversation to a direct message.
## Notification preferences
* Choose [which notifications you receive](https://www.cakewalk.security/docs/human-access/how-to-guides/notification-preferences), by category and by channel, from your profile. This was a top customer request.
* Notifications that require immediate action cannot be switched off.
## No-Touch Onboarding
* Point an integration at [one synced group](https://www.cakewalk.security/docs/human-access/how-to-guides/users-and-groups/joiner-mover-leaver-jml) from your identity provider and new joiners in that group receive their default apps with no review step.
* It combines with Auto-Provisioning, so a joiner can be fully set up without anyone opening Cakewalk.
## Table views persist
* Filters, columns and sorting are stored on your account, so a view is the same on every device and in every tab.
## Non-human identities
* Agents and [service identities](https://www.cakewalk.security/docs/human-access/concepts/data-models/nhi-data-model) found in connected apps appear in their own table, with a compact detail popover.
* App Governance leaves beta.
## Access Sync coverage
* Access Sync reads users and their access from [n8n](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/n8n) and [Datadog](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/datadog).
* User sync for Linear rebuilt.
## Agent Cake coverage
* New apps: Atlassian and [Tableau Cloud](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/tableau).
Improvements
* The profile page is rebuilt around the new notification and preference settings.
* Agent Cake answers inside product knowledge guardrails, so it declines rather than guesses.
* Agent Cake recovers stuck configurations without a support ticket.
* Mixpanel provisioning supports a default project and role.
* Global search is responsive and its input keeps a fixed width.
* Side panels show a proper loading and error state instead of an empty frame.
* Zapier and Calendly handle a seat limit on the target app.
Fixes
* Tables served stale data after a filter change.
* Default tab widths could stick at the wrong size.
* The remove access panel was missing its tooltip.
* An empty User Group table showed no placeholder.
* Remove access requests are scoped correctly for app owners.
* The requested permission level was missing from grant access requests.
* Importing apps from a CSV failed on some files.
* Feedback buttons in the Slack assistant sent an invalid payload.
* Apps that were hidden and set as default raised an error.
{% endupdate %}
{% update date="2026-02-27" tags="new-releases,improvements,fixes" %}
## Bulk access in Slack
Access [requests](https://www.cakewalk.security/docs/human-access/how-to-guides/requests) stop being one app at a time. Agent Cake handles several people and several apps in a single message, and validates a removal list before anything is applied, so a bad list fails before it changes access.
* Cancel a request and search active requests without leaving Slack.
## Non-human identity discovery
* Agents and [service identities](https://www.cakewalk.security/docs/human-access/concepts/data-models/nhi-data-model) discovered in connected apps are collected in one table for security and compliance review.
* Visibility is read only. Requests, Policies and provisioning do not apply to them.
## Access Sync coverage
* Access Sync reads users and their access from [Sentry](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/sentry).
## Agent Cake coverage
* New apps: [Zoom](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/zoom), [Mixpanel](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/mixpanel) and [Robin](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/robin). Mixpanel and [Amplitude](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/amplitude) support custom permission levels.
* More reliable deprovisioning for [Notion](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/notion), [ChatGPT](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/chatgpt), [Amplitude](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/amplitude), [Loom](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/loom), [Productboard](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/productboard), [Figma](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/figma), [Read.ai](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/read.ai), [Pitch](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/pitch) and [Miro](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/miro).
* [Loom](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/loom) rebuilt on browser automation.
Improvements
* Export to JSON.
* Custom fields are available as table columns.
* Filter the users overview by email.
* New progress and loading animations, and skeletons while a table loads.
* Email aliases are readable in the user detail card.
* Agent Cake keeps its session between provisioning runs on many apps, which makes runs faster.
* Agent Cake error handling is centralized, so a failure reports one clear reason instead of a stack trace.
Fixes
* Shared view links were broken and missing their parameters.
* Access Reviews did not advance to the next item.
* Slack sign in failed for workspace owners and when a password picker appeared.
* Confirming removals failed when the offboarding had been cancelled.
* Claiming a standalone task showed the wrong state until refresh.
* Snooze accepted an invalid date.
* The reason field rejected valid input.
* Manage synced groups failed to navigate on close.
* Zendesk, Datadog, Apollo, Miro and Notion sign in and deprovisioning fixes.
{% endupdate %}
{% update date="2026-01-30" tags="new-releases,improvements,fixes" %}
## Tables rebuilt
Every table in Cakewalk gets the same filtering, grouping and scrolling behaviour, so what you learn on one works on the rest.
* New filters across users, user accesses, work app users, [Access Reviews](https://www.cakewalk.security/docs/human-access/how-to-guides/access-reviews), request logs and group imports.
* Infinite scroll with an x of y count, so a long list no longer needs paging.
* Group by moved into a gear menu, with a new interface and a visibility panel beside it.
* Custom fields are available in table queries, including calculated ones.
## Access conflicts
* Conflicting access is recorded and flagged with a warning icon and a tooltip, so an Admin sees the conflict in the table rather than after granting.
## Public API
* New endpoints for work apps, and [API credentials](https://www.cakewalk.security/docs/human-access/open-api-and-mcp/authentication) to call them.
## Access Sync coverage
* Access Sync reads users and their access from [HubSpot](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/hubspot), Salesforce, [PostHog](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/posthog) and [Hex](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/hex-technologies).
## Agent Cake coverage
* New apps: [Lucid](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/lucid), [Typeform](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/typeform), [n8n](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/n8n), [Statsig](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/statsig), [ZenHub](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/zenhub) and [Productboard](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/productboard).
* Deprovisioning added for [Confluent](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/confluent), [Metabase](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/metabase), [Apollo](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/apollo), [Dovetail](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/dovetail), [Intercom](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/intercom), [Docusign](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/docusign) and [Jira](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/jira), so leavers are removed automatically as well as onboarded.
* [Linear](https://www.cakewalk.security/docs/human-access/connections-and-integrations/integration-guides/linear) provisioning rebuilt, with an optional workspace choice.
Improvements
* A request approves itself when the access already exists in the connected app and the data matches.
* The web app loads faster through code splitting.
* Auto-Provisioning keeps its session between runs, which cuts sign in work on every run.
* A failed app integration shows the connection error in detail rather than a generic message.
* Error messages name your company correctly.
* Webhooks send a display name rather than only an identifier.
* Tracked replaces Accepted as the label in the user side panel.
* Germany is the default country on new configurations rather than the United States.
Fixes
* Users could not decline onboarding.
* Claiming a task to add an app left the task stuck.
* Changing the owner of an Access Review failed.
* An offboarding review task ignored access that was already missing.
* App counts in the overview match the underlying list.
* Access Reviews rendered behind other elements.
* A duplicate request reported a conflict instead of pointing at the existing one.
* Jira, DocuSign, Clay, Amplitude, Lovable and Miro sign in fixes.
* Slack reported a wrong password check as a generic failure.
{% endupdate %}
{% endupdates %}
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://www.cakewalk.security/docs/changelog/readme.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.