Changelog

What's new in Cheqpoint

More updates coming weekly. Follow along on X

July 2026v2.2.0New

Internal Notes, Request Watchers, Slack Decisions & Webhook Observability

  • ·Internal notes: leave timestamped comments on any request — assignees and watchers are notified by email instantly (all plans, up to 5 notes per request)
  • ·Request watchers / CC: tag teammates on a request so they receive notifications on all decisions and notes, without being the assigned reviewer (Growth+)
  • ·Slack approve/reject buttons: reviewers can approve or reject directly from the Slack notification — the message updates in place and the decision fires downstream webhooks immediately (all plans with Slack connected)
  • ·Webhook delivery log: Settings → Webhooks tab shows a full log of every outbound webhook — status code, URL, attempt count, and timestamp — so you can diagnose delivery failures without digging through logs (Growth+)
  • ·SDK retryable flag: the approval poll endpoint now returns retryable: true when a request was rejected due to timeout or SLA breach, giving SDK integrations a clear signal to retry cleanly
  • ·Decision payload validation: modified payloads submitted via the decide endpoint are now validated as valid JSON objects or arrays server-side — malformed payloads are rejected with a clear 400 error before any write occurs
  • ·Comment notification emails: internal note notifications now use a dedicated email template with the note body and a direct link to the request, replacing the previous workaround that used the decision email template
  • ·SDK Releases: `@cheqpoint/sdk` 0.1.4 (JS) and `cheqpoint` 0.1.3 (Python) are now available on npm/PyPI, featuring support for the new retryable flag and stability improvements.
July 2026v2.1.0New

Working Hours, Morning Digest & Out-of-Office Delegation

  • ·Business hours SLA pausing (Growth+): configure working days and hours — SLA timers pause overnight and on weekends so deadlines are measured in real working time
  • ·Morning digest: when working hours are enabled, requests that arrive overnight are batched into a single Slack/email notification fired at the start of the next working day
  • ·Per-queue 24/7 override: mark a Review Group as always-on so its SLA timers run on wall-clock time regardless of workspace working hours — ideal for on-call and security queues
  • ·Inbox indicators: off-hours banner and per-card 'Submitted outside working hours' label so reviewers understand timing context at a glance
  • ·Out-of-office delegation (all plans): set a return date and delegate your incoming approvals to a colleague — requests reroute automatically, OOO clears itself when the return date passes
  • ·SDK: processRequest now returns slaAdjustedForWorkingHours: true when the deadline was calculated using working-hours logic
  • ·Audit log: enabling or disabling working hours is recorded with the full schedule snapshot
  • ·Docs: new Working Hours & Out of Office guide at /docs/working-hours
June 2026v2.0.0New

MCP Server — Claude, Cursor & Windsurf Integration

  • ·New @cheqpoint/mcp package: Claude Code, Claude Desktop, Cursor, and Windsurf can now call request_human_approval natively via the Model Context Protocol
  • ·One-line setup: claude mcp add cheqpoint -e CHEQPOINT_CONNECTION_KEY=cq_live_... -- npx -y @cheqpoint/mcp
  • ·get_approval_status tool for polling async requests by ID
  • ·Full error discrimination: APPROVED, REJECTED, TIMED_OUT, and ERROR each return distinct, model-actionable responses
  • ·Configurable timeout via CHEQPOINT_TIMEOUT_MS env var (default 5 min)
June 2026v1.10.1Improved

SDK v0.1.3 (JS) & v0.1.2 (Python) — Framework Fixes

  • ·JS SDK now accepts connectionKey (matching the Cheqpoint dashboard label) in addition to apiKey — both work, with a clear error if neither is provided
  • ·Vercel AI SDK adapter updated for AI SDK v5/v6: inputSchema support, jsonSchema() helper, stopWhen/stepCountIs patterns, toUIMessageStreamResponse()
  • ·LangChain tool updated to import from langchain-core (1.x) with fallback to older langchain package
  • ·Python LangChain and CrewAI tools now distinguish REJECTED vs TIMED_OUT vs ERROR — a network failure no longer reports as a human rejection
  • ·New cheqpoint.langgraph_tool module: explicit LangGraph entry point, works with create_react_agent with no checkpointer required
  • ·Python extras: pip install cheqpoint[langchain], cheqpoint[langgraph], cheqpoint[crewai] now install the correct optional dependencies
  • ·Docs fixed: riskLevel (invalid) replaced with riskScore (0.0–1.0) across all HTTP, inbound, and integration examples
June 2026v1.10.0Improved

Governance Positioning & Homepage

  • ·New Governance section on homepage: auto-decision rules, kill switch, audit export, and EU AI Act Article 14 coverage — above the fold for compliance buyers
  • ·EU AI Act guide linked from homepage and footer Resources column
  • ·Font weight design system enforced across all 25 affected files — no more font-bold violations
  • ·CSS design tokens applied to all hardcoded colours (amber/emerald/red) for correct dark mode behaviour
May 2026v1.9.0Improved

Branding Polish, Search & UX Fixes

  • ·All integration and AI Assistant logos now display in official brand colors (Stripe, Slack, OpenAI, Salesforce, Freshdesk and more)
  • ·Fixed Stripe logo orientation — now renders the correct branded 'S' from the official icon source
  • ·Copy-to-clipboard now shows a toast confirmation for API keys and code snippets
  • ·AI Assistant cards use branded color backgrounds with drop shadows for a premium look
  • ·⌘K Global Search now covers all requests (pending, approved, rejected) with case-insensitive matching
  • ·Last-seen timestamp on AI Assistant cards turns green when an assistant is actively sending requests
  • ·Demo account updated with OpenAI, Salesforce Einstein, and Freshdesk AI Assistants
  • ·Fixed Connection Key copy button in the AI Assistant creation dialog
May 2026v1.8.0Improved

UI Refinement & Platform Stability

  • ·Standardized UI count badges and status markers with circular geometry across the dashboard
  • ·Refined sidebar layout and typography consistency for a more premium experience
  • ·Aligned marketing page CTA buttons for better visual balance on desktop
  • ·Resolved production build stability issues and improved Prisma environment handling
April 2026v1.7.0New

Session Tokens & Kill Switch

  • ·Generate short-lived session tokens scoped to individual AI Assistants (Business+)
  • ·Kill switch lets you instantly pause a specific AI Assistant across all active sessions
  • ·Tokens expire automatically — no manual rotation required
April 2026v1.6.0New

Rule Playground & Risk Policies

  • ·Test auto-decision rules against real historical requests before enabling them (Growth+)
  • ·Risk & Policies page: set workspace-wide risk thresholds and override sensitivity
  • ·Rules engine now shows 'Used X times' so you can spot unused or underperforming rules
April 2026v1.5.0New

Two-Factor Authentication

  • ·TOTP-based 2FA with any authenticator app (Google Authenticator, Authy, 1Password)
  • ·Backup codes generated on setup for account recovery
  • ·Managed from Settings → Security tab
April 2026v1.4.0New

Review Groups with SLA Timers

  • ·Create named review groups and assign specific team members to handle request types
  • ·Set per-group response time targets — requests turn amber then red as SLA approaches
  • ·SLA breach notifications fire to Slack, email, and PagerDuty automatically
April 2026v1.3.1New

AI Assistant Management Dashboard

  • ·Dedicated AI Assistants page to create and manage named Assistants
  • ·Per-Assistant connection keys for fine-grained access control
  • ·Configurable allowed actions, max amount, and hourly/daily rate limits per Assistant
March 2026v1.3.0New

Full Audit Log

  • ·Complete audit trail of all decisions and workspace changes
  • ·Filterable by action type, date range, Assistant, and reviewer
  • ·CSV export ready for compliance reviews
March 2026v1.2.0Improved

Advanced Analytics

  • ·Request volume over time with date range filtering
  • ·Risk distribution charts (low/medium/high breakdown)
  • ·Assistant and reviewer performance tables with average response times
March 2026v1.1.0New

Integrations: Slack, Discord, Teams, Zapier, n8n, PagerDuty

  • ·Real-time notifications on new requests and decisions across all major platforms
  • ·Slack interactive approve/reject — decisions made directly from messages
  • ·PagerDuty on-call escalation for high-risk requests past their SLA
February 2026v1.0.0New

Cheqpoint v1.0 — General Availability

  • ·Human-in-the-loop approval queue for AI Assistants and automations
  • ·REST API + Python and Node.js SDKs for submitting approval requests
  • ·Multi-tenant workspaces with roles: Owner, Admin, Reviewer, Viewer