Changelog
Every new feature, improvement, and fix in Sim, the open-source AI workspace, with release notes straight from GitHub.
v0.9.63
Features
- feat(search): import large permission snapshots in bounded parts
- feat(library): Best AI Agent Platforms for University Student Services and IT
- feat(library): What Do Enterprise AI Agent Platforms Charge for SSO and Audit Logs?
- feat(library): How Do AI Agents Triage Security Alerts With Human Approval?
- feat(library): AI Agent Skills for Business Teams: A 30-Day Learning Guide
Improvements
- improvement(computer-use): name the activation step when an element refuses an advertised action
- improvement(seo): clarify buying guides and migration checks
- improvement(docs): publish the People guide
- improvement(pii): mask texts with many matches in linear time and retry failed chunks three times
- improvement(search): name each Search MCP result's sender, owner or organizer
- improvement(mothership): keep workflow test-run results inside the agent window
- improvement(search): drop unusable citation IDs and title-only previews from Search MCP results
- improvement(landing): add pricing faq and drop retired sdk claims from llms.txt
Bug Fixes
- fix(integrations): hide retired hugging face providers, encode posthog ids, and trim zep final ids
- fix(integrations): keep saved output references valid and encode firecrawl job ids
- fix(computer-use): observe an app's floating and panel-level windows
- fix(integrations): reject invalid kalshi order flags and anchor the arxiv link match
- fix(integrations): close follow-ups from the integration validation audits
- fix(integrations): validate firecrawl, context.dev, apify, bright data, browser use, reducto, pulse and extend
- fix(integrations): validate clerk, crowdstrike, infisical, ketch, onepassword, persona, rootly
- fix(integrations): validate polymarket, kalshi, sftp, ssh, smtp, jupyter and temporal
- fix(integrations): validate spotify, buffer
- fix(integrations): validate quiver, thrive, profound, agentphone, sendblue, linq, quartr
- fix(integrations): validate mongodb, neo4j and upstash
- fix(integrations): validate exa, tavily, serper, linkup, parallel ai, perplexity, jina, duckduckgo, wikipedia and arxiv
- fix(integrations): validate hugging face, elevenlabs, gamma, higgsfield and dspy
- fix(integrations): validate grafana, posthog
- fix(integrations): validate wealthbox, lemlist, emailbison, instantly, attio
- fix(integrations): validate fathom, grain, obsidian and fireflies
- fix(integrations): validate cursor, devin, vercel
- fix(integrations): validate ashby
- fix(integrations): validate google maps, google translate, google bigquery, google books, google appsheet and google pagespeed
- fix(integrations): validate agentmail
- fix(integrations): validate cloudtrail, codepipeline, sqs and ssm
- fix(integrations): validate pinecone, qdrant, elasticsearch and zep
- fix(computer-use): show computer use controls on organization Desktop settings
- fix(pii): describe the mask client's retry budget the way it actually behaves
- fix(landing): anchor the pricing faq section at #faq
- fix(mothership): extend the workflow pickup grace for a view that re-attaches during it
- fix(mothership): cut oversized run errors after projection and collapse long block runs
- fix(lint): use structural placeholders for table input references
- fix(mothership): skip the 30s workflow pickup wait when no chat view is reading
- fix(lint): check table json inputs that hold bare block references
- fix(desktop): show Desktop settings on the organization and account settings surfaces
- fix(search): widen Slack day modifiers so a member's time zone never empties a date window
Other Changes
- test(desktop): stub the computer use availability hook in the prevent-sleep tests
- docs(library): update best-ai-agents-for-it-helpdesk-teams
Features
- feat(library): Zapier Pricing at Scale: What Tasks Really Cost and When AI Agents Are Cheaper
- feat(library): How Do You Build a Governed Workday AI Agent for Employee Onboarding and Offboarding?
- feat(library): 6 Best StackAI Alternatives in 2026 After the Asana Acquisition
- feat(library): Top AI Support Copilots vs Autonomous Support Agents in 2026
- feat(library): How to Pilot an AI Agent Platform Through Security Review
Improvements
- improvement(computer-use): list only what changed in the observation after an action
- improvement(ci): reuse bounded compiler state for acceptance builds
- improvement(ci): omit unused browser maps from acceptance builds
Bug Fixes
- fix(gitlab): scale group discovery and permission-aware search
- fix(ci): restore ESM checks and synchronize docs searches
- fix(billing): withhold hosted Fal.ai image generation from free-plan payers
Other Changes
- ci: bump the actions group across 1 directory with 3 updates
v0.9.61
Improvements
- improvement(comparisons): add langchain migration steps and human approval channels
- improvement(search): describe Slack file hits and leave out GitLab sources that grant the reader nothing
- improvement(ci): scope reviewed root audits to their checks
- improvement(resume): resolve a pause point's block from its loop-scoped context id
Bug Fixes
- fix(search): say a Slack file hit can return its full text rather than promising it
Features
- feat(search): add GitLab group sources with per-project permissions
- feat(library): Best AI Agent Platforms for Refund and Purchase Order Approvals in 2026
- feat(library): How to Replace Brittle Automation Scripts With Reliable AI Agents
Improvements
- perf(sse): split only the newly decoded text in stream line readers
- improvement(library): check sim's integration counts against the catalog
- improvement(computer-use): wait for an app's first observable window after launching it
- improvement(computer-use): report an advertised action the element refuses as unsupported, with the snapshot still valid
Bug Fixes
- fix(library): attribute sim counts by clause and table row, not capitalized words
- fix(search): validate GitLab group scope before permission refresh
- fix(computer-use): find a save panel control's window through its ancestors when it reports none
- fix(pii): tokenize long unbroken runs in linear time so workers stop missing health pings
- fix(computer-use): treat an attribute-unsupported refusal of an advertised action as unsupported
- fix(resume): capture pause block names and keep the resume page off the execution snapshot
- fix(people): answer the cookie consent card in the people e2e so it can't block mobile taps
Other Changes
- docs: match integration counts to the catalog
- docs(library): update best-ai-agent-platforms-hr-employee-questions
Features
- feat(library): AI Agent Identity, Credential Management, and Audit Trails: Enterprise Guide
- feat(library): Snowflake and QuickBooks MCP Servers for AI Agents
- feat(library): Glean vs Dust: Which Enterprise AI Platform Should You Choose?
Improvements
- perf(executor): stop re-parsing the whole structured output on every streamed token
Bug Fixes
- fix(integrations): address release review findings on the sweep
- fix(slack): keep refused file reads from suggesting a reconnect and find the download URL for trigger files
- fix(pii): cap each batched NER pass at one window of characters
- fix(gitlab): clarify permission validation failures
- fix(confluence): index text from Caelor Optics HTML macros
- fix(slack): download files from Slack's download URL and never send the token off its file host
- fix(github): encode owner and repo in request paths
- fix(search): follow Slack's file download redirects without carrying the token off its file host
- fix(table): apply filter and sort builder rules pinned on agent tool rows
- fix(search): read the full text of Slack canvases and files instead of the preview
- fix(integrations): close gaps found in the final sweep audit
- fix(integrations): validate dynamics 365, zoho desk, rippling, okta, sailpoint and snowflake
- fix(blocks): read the mapper operation from the tool-call context in j–z blocks
- fix(blocks): read the mapper operation from the tool-call context in internal and a–i blocks
- fix(blocks): read the mapper operation from the tool-call context in google, microsoft and atlassian blocks
- fix(integrations): validate github, gitlab, linear, slack, pagerduty, incident.io, sentry and datadog
- fix(agent): drop tool-row values left from a previous operation
- fix(brex): leave a blank spend-limit buffer unset
- fix(integrations): validate twilio, sendgrid, resend, mailgun, discord, telegram and whatsapp
- fix(integrations): validate shopify, brex, ramp, zoom, calendly, cal.com, typeform, webflow and wordpress
- fix(blocks): read the mapper operation from the tool-call context in aws and infra blocks
- fix(blocks): read the mapper operation from the tool-call context and ratchet raw reads
- fix(providers): mark retired Nebius models and move the default off one
- fix(docs): normalize path separators in the content build on Windows
- fix(pii): bound NER memory per worker so large texts stop OOM-killing Presidio workers
Other Changes
- docs(library): restore faq entries dropped by recent refreshes
v0.9.58
Features
- feat(library): Microsoft Copilot Studio vs Power Automate: Which Should You Use?
- feat(library): How Do You Build a Governed AI Agent for Microsoft Intune Device Management?
- feat(desktop): quick chat window on a global shortcut
Improvements
- improvement(computer-use): keep an element's id stable across observations
- improvement(egress): route workspace requests through the organization's egress gateway
- improvement(search): log the shape of each search and the source, kind, and size of each read
- improvement(api): log the v2 error code, field, and refusal reason on the request's closing line
- improvement(chat): show background task and scheduled job turns as messages
Bug Fixes
- fix(sandbox): skip interval ticks while the previous one is still in flight
- fix(realtime): keep admin block locks out of write-role graph replaces
- fix(sso): admit callbacks under the provider id the plugin resolved
- fix(copilot): hold callerless runs to the read grant for server tool writes
- fix(sso): refuse SSO callback user writes outside the provider's verified domain
- fix(executor): resume a paused draft run on the graph it paused on
- fix(inbox): require an authenticated sender before an email acts as a workspace member
- fix(helm): reject placeholder secrets and stop shipping them in example values
- fix(webhooks): reject auth that reads an unresolved env-var secret
- fix(workspaces): stop workspace admins from moving the billed account
- fix(mcp): gate public workflow MCP servers on public API governance
- fix(mcp): bound legacy mcp server name and description like the v2 contract
- fix(helm): confine the PII service with its NetworkPolicy whenever pii is enabled
- fix(csp): remove shadowed permissive policy from workflow execute routes
- fix(files): compare Markdown toolbars with their current selection
- fix(computer-use): stop reporting closed menu items past the cap as a truncated state
- fix(files): keep Markdown tables readable while scrolling
- fix(integrations): address release review findings on the integration sweep
- fix(workspaces): hand off a departing member's files without binding every file id
- fix(integrations): resolve agent tool-call operations and align remaining requests with vendor specs
- fix(integrations): address the final release audit findings
- fix(integrations): validate monday, asana, clickup, airtable, notion, supabase, amplitude, box, docusign, and dropbox
- fix(computer-use): read elements reported at infinity without a frame instead of crashing the helper
- fix(integrations): validate zendesk, pipedrive, and intercom against vendor specs
- fix(mailchimp): archive members correctly, add permanent delete, and align requests with the marketing api
- fix(integrations): validate square, quickbooks, oracle fusion, sap, and workday; show the oracle fusion icon
- fix(integrations): validate servicenow, salesforce, and hubspot against vendor docs
- fix(knowledge,files): count knowledge base documents per base and index the modified-order file list
Other Changes
- docs(library): update apache-2-0-vs-fair-code
- docs(library): update can-ai-agents-talk-to-each-other-multi-agent-communication-explained
helm-chart-1.11.20
A Helm chart for Sim - the open-source AI workspace where teams build, deploy, and manage AI agents
v0.9.57
Features
- feat(monaco): add Monaco CRM integration
- feat(computer-use): batch 2-8 actions per call and read app state in one round trip per element
Improvements
- improvement(search): skip the Slack channel lookup on channel-scoped searches and read channel names in any script
- improvement(credentials): hold unsaved keyless subjects so trust policies applied later still match
- improvement(blocks): share the tool-call fallback helper and tidy the Hex/Stripe validation code
- improvement(search): name matching Slack channels and say whether a filtered channel exists
- improvement(search-mcp): log why a batched read returned an error for a document
- improvement(search-mcp): log the field names a rejected tool call's arguments carry
- improvement(computer-use): fold empty layout groups and compact the accessibility tree shown to the model
- improvement(computer-use): remove the activity strip above the chat composer
Bug Fixes
- fix(computer-use): type after a focus click only once the caret is confirmed at the end
- fix(hex): type the block projectVersion output as any and mark category descriptions optional
- fix(credentials): cap the Databricks identity check with a deadline and report no hold for an unheld reconnect subject
- fix(monaco): reject a stage key and a stage ID together on opportunity writes
- fix(slack): let Block Kit messages carry notification text
- fix(desktop-browser): refuse typing behind an unmarked full-screen modal or when no sampled point lands
- fix(computer-use): tighten batch and click-to-focus edge cases from the staging audit
- fix(google-drive): keep the page token an agent passes
- fix(blocks): keep agent-supplied tool args that block mappers remap from dedicated fields
- fix(hex,stripe): validate Hex and Stripe against their API docs
- fix(computer-use): click an editor into focus when Accessibility cannot focus it
- fix(aws-marketplace): check the container license's uncounted usage entitlement so every replica turns Enterprise on
Other Changes
- chore(credentials): move the IAM refusal TSDoc back onto iamFailureMessage
- docs(library): update best-ai-agent-platforms-2026
- docs(library): update best-ai-agents-invoice-processing-finance-reconciliation-2026
- docs(library): update open-source-ai-agent-platforms
- docs(library): update best-ai-agents-for-customer-support-automation
- docs(library): update best-ai-agent-marketplaces-template-libraries
helm-chart-1.11.18
A Helm chart for Sim - the open-source AI workspace where teams build, deploy, and manage AI agents
v0.9.56
Features
- feat(demo): auto-confirm demo bookings that came through the verified form and decline the rest
Improvements
- improvement(desktop-browser): let a page finish rendering before observing an action
- improvement(search): shorter Jira any-term budget and a shared batched read budget
- improvement(executor): run the Agent block on the shared model-fallback walk
- improvement(computer-use): allow every app by default, protect sensitive apps, and make per-app approval optional
- improvement(credentials): resolve agent fallback credentials once, Vanta 401 grace, true refusal codes, headless client-credential modes, and registry cleanups
Bug Fixes
- fix(credentials): replay remembered refusals as credential errors, derive the delegable Google services list, and correct a fallback TSDoc
- fix(credentials): offer keyless Google delegation only for services that can be delegated
- fix(aws-marketplace): confirm a container license lapse on a second re-check before switching Enterprise off
- fix(computer-use): refuse Script Editor, Automator and Shortcuts, which run shell commands outside the Terminal control
- fix(search): check the last shortened read window against its budget
- fix(pii): mask texts over spaCy's length limit and drop encoded files from logs
- fix(aws-marketplace): check the container license with CheckoutLicense so AWS's standard buyer policy works
- fix(docker): ship the realtime and migrations images with runtime dependencies only
- fix(desktop): weak-link ScreenCaptureKit so the Computer Use helper loads on macOS 14 and 15
- fix(docs): copy section permalinks and polish outline navigation
- fix(api): match pause points with jsonb_path_exists so a status-filtered run list stays fast
Other Changes
- chore(pii): drop the over-long text example from the mask client's outage comment
- chore(upgrade): drop the stale Cal.com comment on the sales demo link
- chore(analytics): remove Freebuff ad tracking and the Beluga short-link rewrite

