![\"Monomind\"](\"https://raw.githubusercontent.com/monoes/monomind/main/assets/banner.png\")
\n\n \n
\n An open-source MCP server that extends Claude Code with a codebase knowledge graph, persistent memory, and multi-agent coordination.
\n Apache 2.0 licensed · Your code and memory stay local β see [Trust & Security](#trust--security) for the network calls Monomind does make\n
\n ![\"Docs\"](\"https://img.shields.io/badge/docs-monoes.github.io%2Fmonomind-00D2AA?style=flat-square\"){kind=icon}
![\"npm\"](\"https://img.shields.io/npm/v/monomind?color=%2300D2AA&label=monomind&style=flat-square\")
![\"downloads\"](\"https://img.shields.io/npm/dm/monomind?color=%2310B981&style=flat-square\")
![\"stars\"](\"https://img.shields.io/github/stars/monoes/monomind?color=%23F59E0B&style=flat-square\")
![\"license\"](\"https://img.shields.io/badge/license-Apache_2.0-%238B5CF6?style=flat-square\"){kind=icon}
![\"node](\"https://img.shields.io/badge/node-%3E%3D22.12-339933?style=flat-square&logo=node.js&logoColor=white\"){kind=icon}
\n π’ Orgs Β· \n π Quickstart Β· \n β‘ Mastermind Β· \n 𧩠Agents & Skills Β· \n π Commands Β· \n ποΈ Architecture\n
\n\n---\n\n## What is Monomind?\n\nMonomind is an **open-source CLI and MCP server** that plugs into Claude Code, [OpenCode](https://opencode.ai), Antigravity, Kimi Code, and Codex via the standard [Model Context Protocol](https://modelcontextprotocol.io/). It adds capabilities these assistants don't ship with out of the box:\n\n- **Codebase knowledge graph** β tree-sitter parses your code into a SQLite-backed graph of files, functions, classes, and their relationships. Query imports, callers, and blast radius before making changes.\n- **Persistent memory** β a JSON pattern store with episodic recall that survives across sessions. Agents and orgs share context without re-prompting.\n- **Multi-agent coordination** β in-session, spawn ad-hoc agent teams via Claude Code's Task tool; for persistent background work, `monomind org run` starts a real SDK-backed daemon with policy-gated role agents and a live dashboard.\n- **Agents, skills and picking** β ships 84 pickable agents, 83 skills and 376 Org skills, and you add your own as Markdown files. One index of all of them ranks the best fit for each task; the prompt hook puts it in Claude's context as a `[PICK]` line, and `monomind pick` or the `pick` MCP tool return it on request. See [Agents & Skills](doc/concepts/agents-and-skills.md) and [Routing](doc/concepts/routing.md).\n- **Reusable slash commands** β 42 workflows (plan, execute, review, debug, release, research, worktree) available as `/mastermind:*` commands inside Claude Code.\n\n```bash\nnpm install -g monomind # Apache 2.0 licensed, runs entirely on your machine\ncd your-project && monomind init\nclaude mcp add monomind -- npx -y monomind@latest mcp start\n```\n\n| rules | rollback\n```\n\n`org` has 36 subcommands total (skills, run, stop, pause, resume, reload, status, serve, supervisor, test-loop, logs, events, watch, report, memory, costs, inbox, flow, questions, approvals, answer, approve, deny, gates, gate-approve, gate-reject, replay, resume-from, branch, decisions, create, validate, migrate, list, delete, mark-complete).\n\n> **Note:** `/mastermind:runorg` delegates directly to the Org Runtime daemon (the same path as `monomind org run`) β there is no boss agent, no monotask board, and no manual curl calls in this path. `/mastermind:runorg` converts legacy-format org config files with `monomind org migrate` before starting the daemon. New orgs should use `monomind org run` (or `/mastermind:runorg`) against a hand-authored `.monomind/orgs/ .json`.\n\n---\n\n## β‘ Looping Workflows\n\nFor code, the `/mastermind:*` workflows can loop instead of running once. Add `--tillend` to `/mastermind:review`, `/mastermind:improve`, `/mastermind:debug` and most other workflows (`/mastermind:help` lists them) to repeat until a round finds nothing left to do.\n\n```bash\n/mastermind:review --tillend # review β fix β verify until a round finds nothing\n/mastermind:improve --repeat 3 the auth flow # exactly 3 improvement passes\n/mastermind:plan add rate limiting # then /mastermind:execute the plan\n```\n\n### Loop flags\n\n| Flag | Purpose |\n|---|---|\n| `--tillend` | Repeat until an empty round (zero findings, zero actions) |\n| `--repeat ` | Repeat exactly N times |\n| `--maxruns ` | Safety cap for `--tillend` (default 50) |\n| `--wait ` | Pause between runs |\n\n---\n\n## π Quickstart\n\n```bash\n# 1. Install\nnpm install -g monomind\n\n# 2. Initialize in your project\ncd your-project\nmonomind init\n\n# 3. Wire into Claude Code as an MCP server\nclaude mcp add monomind -- npx -y monomind@latest mcp start\n\n# 4. Health check\nmonomind doctor --fix\n```\n\n> **Semantic routing (opt-in download):** embedding-based task routing needs a local model (~88 MB, `Snowflake/snowflake-arctic-embed-xs` via transformers.js). `monomind init` asks interactively whether to download it β the default is No, and non-interactive/CI installs never download it silently. Declining is fine: agent picking (`monomind pick`, the `[PICK]` hook line, `hooks route`) never needs the model; only `route semantic`, `hooks_route_semantic` and `agent spawn --task` use it, after the picker, and fall back to keyword and hash matching without it. Fetch it any time with `monomind download-embeddings` (or `node scripts/download-embedding-model.mjs` on a source checkout).\n\n> **Native module install blocked?** If `doctor` reports a missing `better-sqlite3` binding (`Could not locate the bindings file`, or npm logs an install script that was \"blocked because it is not covered by allowScripts\"), your npm's `allowScripts` policy blocked its native build β this isn't a Monomind bug. Run `npm install-scripts approve better-sqlite3 && npm rebuild better-sqlite3`, then re-run `monomind doctor --fix`.\n\nOpen Claude Code. You now have 42 `/mastermind:*` workflows available:\n\n```bash\n/mastermind:review --tillend # review and fix until nothing is left\nmonomind org run sample-team # run your first AI org (init writes a runnable sample-team.json β edit it, or run it as-is)\n/mastermind:help # show all commands\n```\n\n---\n\n## π Second Brain β Your Documents, Retrieved by Meaning\n\nDrop documents (Markdown, TXT, PDF, DOCX) anywhere in your project and run `monomind init` β the Second Brain activates itself. No flags, no configuration, no accounts. Everything runs on your machine: a local embedding model (`Alibaba-NLP/gte-modernbert-base`, 768-dim, via transformers.js) and a local SQLite vector store. **Your notes never leave your computer.**\n\nFrom then on, every substantive prompt you type in Claude Code is automatically answered *with your own knowledge in context* β a hook retrieves the most relevant excerpts semantically (the always-on dashboard keeps the model warm, ~60ms per lookup) and injects them before Claude starts thinking. Ask \"when do new parents get time off\" and the parental-leave section of your handbook is already on the table, even though you never used the word \"leave\".\n\n```bash\nmonomind doc ingest ./notes # index documents (init + session-start do this automatically)\nmonomind doc search -q \"pricing psychology in checkout\" # semantic search, by meaning not keywords\nmonomind doc list # what's indexed\nmonomind doc export # portable OKF bundle β move your brain between machines\n```\n\n> **Optional spreadsheet support:** `monomind init` never downloads SheetJS. To extract `.xlsx`, `.xls`, or `.ods` files, install it only when you need it: run `pnpm add xlsx` in a project using a local/npx Monomind install, or `npm install -g xlsx` when Monomind is installed globally. Until then, spreadsheet files are skipped while the rest of document ingestion continues.\n\n**And it follows you across projects.** Ingest a path from *outside* the current project (`monomind doc ingest ~/notes`, or add `--global`) and it lands in your personal global brain at `~/.monomind/global-brain` (override with `MONOMIND_GLOBAL_BRAIN_DIR`) β kept as a sibling of `~/.monomind/projects` specifically so `monomind cleanup --data` can never prune it β searchable from every project on the machine. All retrieval (CLI search, per-prompt injection, the dashboard) merges both stores automatically, with project knowledge winning ties and global hits labeled `[global]`. `doc export --global` moves your whole brain between machines as an OKF bundle β still no cloud, ever.\n\nRetrieval quality is a tested invariant, not a hope: a golden-set eval (paraphrase queries against notes written in different vocabulary) runs in CI with an 80% recall bar.\n\n> **Privacy note:** the embedding model (~90MB) is fetched once from HuggingFace's CDN when your first document is indexed, then cached locally forever. That download is the only outbound request the Second Brain ever makes β your documents and queries never leave your machine. Offline at first index? Search degrades gracefully to keyword matching and `monomind doctor` tells you how to warm up later.\n\n> **Which model where?** Monomind uses two local embedding models. `Snowflake/snowflake-arctic-embed-xs` (~88MB) serves semantic task routing only. Everything the memory bridge embeds β both the Second Brain document index and the persistent memory store, which share that one bridge β uses `Alibaba-NLP/gte-modernbert-base` (768-dim, ~90MB); there is no separate document model. See [Embeddings](./doc/commands/memory.md#hybrid-search-architecture--options) for the per-subsystem detail.\n\n---\n\n## π§ Memory That Persists\n\nEvery session, every agent, every org writes to a persistent memory store that survives across sessions β text plus embedding vectors in local SQLite (better-sqlite3, pure-WASM fallback), embedded by a local model. No cloud vector database, no API keys, no data transmission. The next time you run anything, Monomind already knows what was built, what failed, and which patterns work.\n\n```mermaid\ngraph TD\n L0[\"L0 - In-flight\\nCurrent session drawers\\nephemeral\"]\n L1[\"L1 - Working\\nCross-session memory\\nBM25 K1=1.5, B=0.75\"]\n L2[\"L2 - Long-term\\nEpisodic store\\nSemantic recall\"]\n L3[\"L3 - Shared\\nCross-agent namespace\\nFederated swarm reads\"]\n\n L0 -->|promoted| L1 --> L2 --> L3\n\n style L0 fill:#00D2AA11,stroke:#00D2AA\n style L1 fill:#F59E0B11,stroke:#F59E0B\n style L2 fill:#8B5CF611,stroke:#8B5CF6\n style L3 fill:#EF444411,stroke:#EF4444\n```\n\n```bash\nmonomind memory store --key \"key insight\" --value \"β¦\" --namespace my-project\nmonomind memory search \"auth implementation\" # semantic (local embeddings) with keyword fallback\n```\n\n---\n\n## πΊοΈ Monograph β Your Codebase, as a Graph\n\nBefore touching any file, Monomind queries **Monograph** β a SQLite-backed knowledge graph of your entire codebase. Nodes are files, classes, and functions. Edges are imports, calls, and dependencies.\n\n```bash\n/mastermind:understand # build the graph\n/mastermind:graph-status # nodes Β· edges Β· freshness\n\n# Inside Claude Code, Monograph runs automatically:\n# β \"what files does auth.ts import?\"\n# β \"what breaks if I change UserService?\"\n# β \"find all callers of validateToken()\"\n```\n\n19 default MCP tools (+27 advanced via `MONOGRAPH_MCP_ADVANCED=1`). Impact analysis. Community detection. Zero grep.\n\n---\n\n## π£ Hooks & Workers\n\nMonomind wires 28 hook subcommands into Claude Code across edit, task, command, and session lifecycle events β logging patterns, routing agents, and feeding the intelligence system.\n\n```mermaid\nflowchart LR\n CE[\"Claude Code\\nEvent\"] --> H[\"Hook Router\"]\n H --> P[\"pre-edit\\npre-task\\npre-command\"]\n H --> SS[\"session-start\\nsession-end\\nnotify\"]\n H --> I[\"route\\nlearn\"]\n H --> T[\"teammate-idle\\ntask-completed\"]\n\n I --> DB[(\"patterns.json\\nmemory store\")]\n DB -->|next session| CE\n```\n\n**9 on-demand workers** run at session start (staleness-gated, refreshed when older than 6 hours): `health` Β· `ddd` Β· `security` Β· `cache` Β· `progress` Β· `map` Β· `audit` Β· `consolidate` Β· `reflexion`.\n\n---\n\n## π‘οΈ MonoFence AI β Security Layer\n\nEvery agent boundary is defended by **monofence-ai** β real-time detection of prompt injection, jailbreaks, homoglyphs, base64 evasion, multi-turn escalation, and PII leakage.\n\n```typescript\nimport { isSafe, createMonoDefence } from 'monofence-ai';\n\nisSafe('Ignore all previous instructions'); // β false (~0.04ms)\n\nconst fence = createMonoDefence({ enableContextTracking: true });\nconst result = await fence.detect(userInput);\n// result.safe Β· result.threats Β· result.overallRisk\n```\n\nIn Claude Code, the live pre-bash/pre-write gate is wired up via its own lazy-loaded integration in `.claude/helpers/handlers/gates-handler.cjs` (`MONOMIND_MONOFENCE_GATE=off` to disable) β not via monofence-ai's `registerSecurityHooks()` API, which is a separate integration point consumed only by `@monoes/hooks`' in-process `HookExecutor`.\n\n---\n\n## π 42 Mastermind Commands\n\nEverything runs from inside Claude Code via slash commands. Here's the highlight reel:\n\n### Development\n| Command | What it does |\n|---|---|\n| `/mastermind:plan` | Comprehensive implementation plan |\n| `/mastermind:execute` | Execute a written plan step by step, then hand off to review |\n| `/mastermind:review` | Iterative review until zero findings |\n| `/mastermind:debug` | Systematic root-cause debugging |\n| `/mastermind:improve` | Analyze a component and write improvement tasks |\n| `/mastermind:release` | Versioning, changelog, deployment coordination |\n| `/mastermind:worktree` | Feature work in isolated git worktree |\n\n### Organizations\n| Command | What it does |\n|---|---|\n| `monomind org run ` | Start an org as a real SDK-backed daemon |\n| `monomind org status` / `list` | Runtime state for one or all orgs |\n| `monomind org stop ` | Request a graceful stop |\n| `monomind org approve ` / `deny` | Act on pending approval requests |\n\n### Business Domains\n| Command | What it does |\n|---|---|\n| `/mastermind:marketing` | Campaigns, copy, SEO, social |\n| `/mastermind:content` | Blog posts, threads, newsletters |\n| `/mastermind:sales` | Outreach, proposals, pipeline |\n| `/mastermind:finance` | Budgets, invoicing, modeling |\n| `/mastermind:ops` | Operations and workflow automation |\n\n**[β Full reference (42 commands)](https://monoes.github.io/monomind/#slash)**\n\n---\n\n## π¦ Packages\n\n| Package | npm | Purpose |\n|---|---|---|\n| `monomind` | [](https://www.npmjs.com/package/monomind) | Umbrella shim β **install this one** |\n| `@monoes/monomindcli` | [](https://www.npmjs.com/package/@monoes/monomindcli) | CLI engine (38 commands, MCP server) |\n| `@monoes/monograph` | [](https://www.npmjs.com/package/@monoes/monograph) | Code knowledge graph (tree-sitter + SQLite) |\n| `@monoes/memory` | [](https://www.npmjs.com/package/@monoes/memory) | Persistent memory backends (SQLite + vectors) |\n| `@monoes/hooks` | [](https://www.npmjs.com/package/@monoes/hooks) | Hook registry + 9 on-demand workers |\n| `@monoes/mcp` | [](https://www.npmjs.com/package/@monoes/mcp) | MCP server framework (stdio/HTTP/WebSocket) |\n| `@monoes/routing` | [](https://www.npmjs.com/package/@monoes/routing) | Semantic task-to-agent routing |\n| `@monoes/monobrowse` | [](https://www.npmjs.com/package/@monoes/monobrowse) | Browser automation via CDP |\n| `@monoes/monodesign` | [](https://www.npmjs.com/package/@monoes/monodesign) | Frontend design intelligence |\n| `monofence-ai` | [](https://www.npmjs.com/package/monofence-ai) | AI manipulation defence |\n\nSee [CLI Reference](./doc/commands/cli-reference.md) for the full 38-command index.\n\n---\n\n## ποΈ How It's Built\n\n```mermaid\ngraph TD\n CC[\"Claude Code\"]\n MCP[\"MCP Server\\nmonomind mcp start\"]\n D[\"Background Workers\\n(@monoes/hooks, in-process)\"]\n ORG[\"OrgDaemon\\nmonomind org run\\nreal SDK sessions\"]\n\n CC <-->|\"MCP tools: monograph, memory\"| MCP\n MCP <--> D\n\n D --> ADB[(\"Memory store\\npatterns + episodes\")]\n D --> MG[(\"Monograph\\ncode graph\")]\n D --> HK[\"Hooks\\n28 subcommands\"]\n\n CC -->|\"Task tool - spawns agents\"| AG[\"In-session agents\\narchitect, coder\\ntester, reviewer\\nsecurity, perf\"]\n AG <-->|reads and writes| ADB\n ORG -->|\"spawns, policy-gated\"| RA[\"Role agents\"]\n ORG <-->|reads and writes| ADB\n\n style CC fill:#00D2AA22,stroke:#00D2AA\n style AG fill:#8B5CF622,stroke:#8B5CF6\n style ADB fill:#F59E0B22,stroke:#F59E0B\n style ORG fill:#F59E0B22,stroke:#F59E0B\n```\n\n**Claude Code's Task tool drives in-session multi-agent work; `monomind org run` drives persistent background orgs.** Your code and memory stay local β see [Trust & Security](#trust--security) for the network calls Monomind does make.\n\n---\n\n## Resources\n\n- π [Full Documentation](https://monoes.github.io/monomind/)\n- π₯οΈ [CLI Command Reference](https://github.com/monoes/monomind/blob/main/doc/commands/cli-reference.md)\n- π’ [Org Runtime Architecture](https://github.com/monoes/monomind/blob/main/doc/concepts/org-runtime.md)\n- π’ [Autonomous Orgs](https://monoes.github.io/monomind/#orgs)\n- β‘ [Mastermind Reference](https://monoes.github.io/monomind/#mastermind)\n- π [All Slash Commands](https://monoes.github.io/monomind/#slash)\n- π [Issues](https://github.com/monoes/monomind/issues)\n- π¬ [Discussions](https://github.com/monoes/monomind/discussions)\n\n---\n\n \n
\n","readmeFilename":"README.md"}
\n Built with β₯ by monoes Β· Apache 2.0 License\n
![\"Monomind\"](\"https://raw.githubusercontent.com/monoes/monomind/main/assets/mascot.png\")