Studio now writes down what changed on a project, who changed it, and what it looked like before.
The project audit log
Every create, update, and delete is recorded against the entity it touched — stories, epics, bugs, sprints, requirements, wiki pages and folders, canvas diagrams, resources, meetings, test suites, cases, plans and environments, project members, API keys, comments, and project context. The log lives on the project and can be filtered by entity type, action, and actor.
Actors are recorded by kind, not just by name. A change made by a teammate, by Helix, by an MCP client, or by a scheduled system job each carries its own label, so an unexpected edit is traceable to the thing that made it rather than to whoever happened to own the session.
Sensitive values — API keys, tokens, credentials — are redacted before the entry is written, and the log entry is committed in the same transaction as the change itself. An action that fails leaves no audit row, and an action that succeeds cannot silently skip one.
Also in this release
- Universal integration read endpoints. External tooling can read the core project entities through a single consistent set of endpoints instead of reaching for per-feature APIs.
- Scoped MCP toolsets. A
toolsetsquery parameter lets an MCP client load only the tool groups it needs rather than all of them, which keeps smaller clients well inside their tool limits. - Setup guidance for claude.ai and ChatGPT in the MCP docs, with client pages now linking straight to toolset configuration.
- Docs pages support Loom embeds and shareable tab URLs, so a link can point at the exact tab you were reading.