VERSION CLUSTER (v3.6 5/29 + v3.7 early June, all vendor changelog): **3.6 'Auto-review Run Mode'** (5/29) -- a classifier subagent routes Shell/MCP/Fetch tool calls between auto-execute and require-review, cutting approval fatigue without going full-YOLO (the aggregator-circulated 'Cursor 3.6 auto-review' claim is confirmed real). **v3.7 ships**: Organizations for Cursor Enterprise (6/3, per-team security/governance/budget controls), Canvas Design Mode + Context Usage Report (6/4), SDK upgrades (custom stores, custom tools, auto-review, nested subagents in TS/Python), and Design Mode improvements (6/5, multi-select + voice input). Enterprise governance + design tooling in one week -- Cursor pushing both upmarket and into the designer workflow simultaneously
2026-06-05•Cursor changelog (cursor.com/changelog) -- 2026-05-29 v3.6 + 2026-06-03/04/05 v3.7 entries
PRODUCT (2026-05-19 + 2026-05-20): Cursor shipped two post-Composer-2.5 integration/feature drops. **5/19 Cursor in Jira**: @Cursor mentions inside Jira comments trigger cloud agent runs that auto-create PR links back to the Jira ticket. Requires Rovo-enabled Jira Commercial Cloud (Atlassian's AI tier) -- not available on Jira Server / Data Center. **5/20 Cursor Automations v2**: multi-repo support, no-repo automations (automations that orchestrate without targeting a specific repo at definition time), plus 5 new marketplace templates (Slack digest agent, product analytics agent, FAQ agent, finance agent, customer-health agent). **50% discount on agent runs for the first 7 days of each new automation**. The Jira surface adds @-mention parity with the 5/11 Microsoft Teams integration; the Automations v2 marketplace template push is Cursor's bid for the no/low-code business-agent slot that Microsoft Agent 365 + Notion Custom Agents are competing for.
2026-05-20•Cursor changelog (cursor.com/changelog) -- 2026-05-19 Jira + 2026-05-20 Automations v2
PRODUCT (2026-05-18): **Cursor Composer 2.5** shipped via the Cursor changelog -- 'substantial improvement in intelligence and behavior' over Composer 2, with the **same pricing** structure ($0.50/M input + $2.50/M output for the Standard tier; Fast tier remains $1.50/$7.50 for Pro+ subscribers). **Double usage for the first launch week** -- effectively a free quality bump for all paying users through ~5/25. Vendor changelog confirmed at cursor.com/changelog; the aggregator-circulated 'Kimi K2.5 architecture' detail is not stated in the official changelog post and should be treated as unverified narrative. Practical implication: if you were holding Composer use down because Composer 2 felt unreliable on architectural tasks, retry with Composer 2.5 this week. Pair with the 5/13 cloud Dev Environments + 5/7 PR review experience for the strongest one-month feature window since Cursor 3 launched.
2026-05-18•Cursor changelog (cursor.com/changelog) -- 2026-05-18 Composer 2.5
PRODUCT (2026-05-07): **Cursor 3.3** shipped with a new PR review experience (Reviews / Commits / Changes tabs inside the editor) and **'Build in Parallel'** -- Cursor multitasks plan execution so the agent works through multiple independent steps simultaneously instead of serializing them. Pinnable quick-action skill pills also landed. The PR review surface is a clear shot at GitHub Copilot Code Review's territory (Copilot is also re-pricing PR review on Actions minutes 6/1, see github-copilot.ts) -- Cursor's version sits in the editor and doesn't bill separately.
2026-05-07•Cursor changelog (cursor.com/changelog/05-07-26)
PRODUCT (2026-05-13): Cursor shipped 'Development Environments for Cloud Agents' -- configured runtime setups so cloud agents can take engineering tasks start to finish with cloned repos, installed deps, internal-toolchain credentials, and build-system access. Key capabilities: (a) multi-repo environments (one agent works across interdependent codebases), (b) Dockerfile-based config with build secrets and improved layer caching (70% faster builds), (c) Cursor-generated Dockerfiles via agent-led setup (private beta for Enterprise -- agent inspects repo, asks clarifying questions, validates), (d) governance controls: version history with rollback, audit log of all changes, granular permissions (secrets + network egress scoped per environment). Positions Cursor's cloud agents as a viable Devin / Cognition substitute for teams that want the agent inside their existing Cursor seat instead of buying a separate Cognition contract.
2026-05-13•Cursor blog (cursor.com/blog/cloud-agent-development-environments), Cursor changelog 2026-05-13
PRODUCT (2026-05-11): Cursor in Microsoft Teams -- @Cursor mentions inside Teams channels delegate a coding task to the Cursor cloud agent, which auto-selects the relevant repo + model, reads thread context, and opens a PR. First first-party Cursor integration outside the editor itself; positions Cursor against GitHub Copilot's Teams-native presence. Concurrently shipped: Bugbot effort levels (Default / High / Custom tiers for code-review depth -- High pass tradeoffs more deeply, Custom lets you set the budget per repo).
2026-05-11•Cursor changelog (cursor.com/changelog/microsoft-teams), Cursor changelog (cursor.com/changelog)
Issues here are sourced from our editorial sweeps, not real-time telemetry. Newer issues may exist.