**When a coding agent dies mid-task, the task frees itself.**
以下摘录自该仓库的 README(片段) · 在 GitHub 读完整 README →
**rhizome-mcp** gives autonomous coding agents crash-safe task coordination over MCP: claims are renewable expiring leases, in_progress is derived — never stored — and an interrupted attempt hands its checkpoint to whichever session picks the work up next. One static Go binary, one SQLite database per project. No daemon, no accounts, no cloud.
It works with agents from different products at once — Claude Code, Codex, GitHub Copilot, VS Code, and any other MCP-compatible client — giving them one shared, durable view of project work.
Recorded from real server output by demo/record.sh — nothing staged.
**Why** · **How it compares** · **Quick start** · **Monitor your project** · **MCP surface** · **Documentation**
Why
AI coding agents are concurrent, context-limited, and interruptible. A TODO.md or a single chat context doesn't survive that. rhizome-mcp is built around those failure modes:
- **Crash-safe claiming.** Issues are claimed atomically with renewable leases. in_progress is never a stored status — it is derived from an active lease, so a vanished agent can't lock an issue forever. When the lease expires, the issue becomes claimable again. A partial unique index guarantees at most one active attempt per issue at the database level.
- **Token-efficient by contract.** Compact list projections (a 100-issue page stays under 64 KB — enforced by an integration test), graph nodes that exclude free-text bodies at the SQL layer, snippet-only search, delta sync via event IDs, and a bounded single-call work-context package.
- **Durable project memory.** Checkpoints with next steps, supersedable decision records, append-only event history, and FTS5 full-text search across issues, comments, decisions, and notes. A fresh session resumes from the last checkpoint instead of re-deriving state.
- **Planning and dependency graphs.** Cycle-checked blocks relations, epics, claimable entry-point highlighting, and atomic batch planning (up to 50 issues, 100 relations, and 20 decisions in one all-or-nothing transaction).
- **Review workflow.** Review requests pin an exact issue version and event position; approving changed code is structurally impossible — a stale request can only be superseded and re-pinned.
Watch a stale approval get refused
- **Resource reservations.** A claim can atomically reserve files, directories, globs, or logical resources (a port, a migration window, a deploy slot); an overlapping claim fails fast with the holder and its lease expiry named, instead of two agents colliding later.
Watch a reservation conflict fail a claim atomically
- **Concurrency discipline throughout.** Optimistic versioning on mutations, replay-safe idempotency keys, stable machine-actionable error codes.
- **Human observability without a server.** rhizome-mcp board prints live leases, blockers, and the review queue, or writes a self-contained HTML snapshot; the CLI reads everything as tables, JSON, or Mermaid.
This repository tracks its own backlog through the server it ships — work is selected, claimed, checkpointed, and reviewed via rhizome-mcp itself (AGENTS.md).
**Use it when** several agent sessions (or several agent products) work the same repository over time and you need handoffs, parallel work, and recovery after crashes or context limits.
**Skip it if** you need a hosted multi-user tracker with auth, permissions, and a web UI — this is a local single-developer tool by design.
How it compares
Compare agent task trackers on guarantees under failure, not on feature lists — local-first SQLite storage and MCP support are table stakes in this category.
| When things go wrong | rhizome-mcp | beads | Kata | Guild |
| --- | --- | --- | --- | --- |
| Task frees itself after a crash | **Yes** — expiring lease | No | No | No |
| Double-claim prevented at the storage layer | **Yes** | Atomic claim | Atomic claim | Atomic claim |
| Stale review approval impossible | **Yes** — version-pinned | No | No | No |
| Response sizes bounded by a tested contract | **Yes** — ≤ 64 KiB / 100 issues | No | No | No |
| Interrupted attempt resumable by another session | **Yes** — checkpoints | No | No | Note only |