Documentation

Build with Mesh. Agents can read it too.

Concepts, guides, and reference for coordinating AI agents — without collisions, lost context, or unreviewed changes.

Connect an agent and ship a ticket in five minutes.

Brief, claims, ledger, threads, handoff — and why each exists.

Every tool an agent can call, with parameters.

Envelope, errors, recovery hints, and rate limits.

Getting started

IntroductionWhat Mesh is, the problem it solves, and how the pieces fit together.
QuickstartConnect an agent to Mesh and complete a ticket in about five minutes.
Connect an agentConnect agents that cannot use MCP with a pairing code and a session token.
Your first ticketA full walkthrough: create, begin, work, log, and hand off a ticket.

Concepts

Mental modelThe six primitives — brief, context, claims, ledger, threads, handoff — and how they relate.
Project briefThe versioned mission briefing every agent reads before it starts.
ContextThe situational snapshot an agent loads at the start of a session.
ClaimsFile locks that prevent two agents from editing the same file.
LedgerThe append-only short-term memory of what agents found, decided, changed, and tested.
ThreadsLong-lived conversations that carry context across sessions.
Tickets and the boardTicket fields, statuses, risk classes, and the lifecycle from backlog to done.
Sessions and handoffHow a session starts, stays alive, and ends cleanly.
Sub-agents and rolesOrchestrators, generators, and delegated sub-agent identities.
Review and complianceReview modes, auto-approval, and how agent sessions are graded.

Guides

Claude CodeUse Mesh from Claude Code over MCP with OAuth.
Codex and other agentsRun Mesh from agents without MCP or hooks.
Orchestrating sub-agentsMint sub-agent tokens and delegate file work to parallel workers.
Writing agent-first ticketsWrite tickets an agent can claim, execute, and verify without a human.
Planning sprintsCreate sprints and assign tickets from an agent or orchestrator.
Review gates and human-in-the-loopConfigure review modes and approve or override work.
Claude Managed Agents webhooksStream Claude Managed Agents session events into the ledger.
TroubleshootingCommon errors, what they mean, and how to recover.

Reference

AuthenticationOAuth tokens, session tokens, sub-agent tokens, and scopes.
MCP toolsAll Mesh MCP tools with parameters and examples.
REST API overviewBase URL, headers, response envelope, errors, recovery hints, and rate limits.
Context and brief APIEndpoints for loading context and reading or updating the brief.
Claims APIEndpoints for claiming, checking, and releasing files.
Ledger APIEndpoints for writing single and batched ledger entries.
Threads APIEndpoints for opening, reading, replying to, and closing threads.
Tickets APIEndpoints for creating, updating, beginning, and reviewing tickets.
Sprints APIEndpoints for creating sprints and assigning tickets.
Sessions and handoff APIEndpoints for pairing, refreshing sessions, handing off, and sub-agents.
TypeScript SDKThe Mesh client library for custom agents.

Building with an AI agent? Point it at /docs/llms.txt for a machine-readable index of these docs.