Platform
Update

Your workspace's tasks, approvals, audit log, usage and skills, from any MCP client

TL;DR

Skills, memory, schedules, alerts, dashboards and assets were already reachable over MCP. What only the REST API answered is now an MCP tool too — list_tasks, list_write_requests, get_audit_log, get_usage and get_organization — each calling the REST endpoint's own reader with the same scopes. And export_skills hands a local agent the workspace's own enabled skills as SKILL.md files, ready for Claude Code or Codex.

The new tools

ToolWhat it answersWho can call it
list_tasksThe workspace's tasks, newest first, filtered by status or by a project's runsA key with tasks:read
list_write_requestsThe writes this key queued for a person, each with its review linkAny key with a tools or tasks scope
get_audit_logWhat this key called, queued and started, and what a person decided — ids, names and outcomes onlyAny key
get_usageCredits this key's calls used, per day and per toolAny key
get_organizationThe workspace's name, plan, credits left and connected accounts the key may seeAny key
export_skillsThe workspace's own enabled skills as SKILL.md files (front matter + playbook)A key with a tools or tasks scope

Load the team's skills in a local agent

  1. 1Ask your agent to call export_skills (optionally with the slugs you want).
  2. 2Each answer names where the file goes: .claude/skills/<slug>/SKILL.md for Claude Code, ~/.codex/skills/<slug>/SKILL.md for Codex.
  3. 3Changed a skill locally? Send it back with internal__sync_skills — a skill someone else changed meanwhile comes back as a conflict, never overwritten.
Built-in playbooks are not exported (they live on the server). A key narrowed to some platforms is offered no skill about a platform it does not expose. Setup: MCP setup guide.

Frequently asked questions

Can Claude Code use my team's Agent Planners skills locally?
Yes. export_skills returns the workspace's own enabled skills as SKILL.md files for .claude/skills, and internal__sync_skills sends changes back.
Can an MCP client read the audit log?
It reads the rows about its own key: what it called, queued and started and what a person decided on its requests — never arguments, bodies or tokens.
Try it on your own account

Start free — 2,500 credits a month, no credit card. Reads run at once, and writes wait for your approval by default.

Related releases