{ "markdown": "# Munin\n\n> Open-source, headless HubSpot alternative.\n\n

\n \"MIT\n \"Last\n \"Listed\n

\n\n

\n Website ·\n See it in action ·\n Documentation ·\n MCP Registry\n

\n\nCRM, conversations, outreach, CMS, knowledge base, and analytics on one Postgres schema — exposed as tools your agents drive, not screens you click through. Headless the way a headless CMS is: there's a thin dashboard for settings, auth, and human-in-the-loop review, but the apps themselves have no admin UI. Every action runs through MCP tools, callable from any MCP-compatible client (Claude, Cursor, ChatGPT, custom runners) — same tools, same permissions, same audit log, whether a human or an agent is driving. Munin even ships its own: an in-process, per-org agent runner that answers live conversations and works the curation queue against an LLM provider you configure — so the platform runs out of the box, with external MCP clients optional.\n\n

\n \"The
\n The dashboard — a thin shell for settings, auth, and human-in-the-loop review. No admin UI for app data; it drives the same MCP tools your agents call.\n

\n\n

\n \"The
\n The embeddable chat widget — answering a live customer from the knowledge base, ready to hand off to a human and be picked back up by the agent.\n

\n\n## Modules at a glance\n\n| Module | Tools | What it does |\n|---|---|---|\n| Knowledge Base | `kb_*` | documents, hybrid search, audience scoping |\n| Conversations | `conv_*` | channels, messages, handover |\n| CRM | `crm_*` | contacts, companies, deals |\n| CMS | `cms_*` | collections, entries, assets |\n| Outreach | `outreach_*` | campaigns, drafts, propose-only |\n| Analytics | `analytics_*` | page-view + search events |\n\nThese six modules aren't separate products — they share one Postgres schema, one permission model, and one audit log. Watch how they tie together:\n\n

\n \n \"Watch:\n \n

\n\n## Core modules\n\n#### Knowledge Base\n- Markdown articles organized into spaces, each scoped to the audiences allowed to see it.\n- Hybrid search that blends keyword matching with meaning-based results.\n- Website import — crawl a public site and turn each page into an article in the background, automatically dropping articles when their source page disappears.\n- Full version history with restore, plus a review queue for proposed edits.\n\n#### Conversations\n- One inbox across email, chat widget, voice (Threll.ai / Vapi), and SMS (Twilio / MessageBird).\n- Inbound *and* outbound — agents answer conversations and can place outbound calls.\n- Assignable, organized by topic, and searchable across every message.\n- Built-in handoff to a human, with notifications to your own systems as conversations change.\n\n#### CRM\n- Contacts, companies, deals, activities, pipelines, and segments.\n- AI-written summaries and suggested next actions, kept separate from what people edit by hand.\n- Consent tracking — the lawful basis and source for each contact, required before they can be added to any outreach.\n- Automatic duplicate detection that proposes merges for review, plus bulk contact import.\n\n#### CMS\n- Content collections with structured fields, and entries you can publish in multiple languages.\n- Rich content blocks for article bodies — callouts, quotes, media, and more.\n- Scheduled publishing and a media library for images and files.\n- Full version history with restore, search, and cross-references between entries.\n- A public content API serves your site or app, with engagement tracking built into every entry.\n\n#### Outreach\n- Propose-only outbound email — campaigns, segments, and drafts for both first touches and replies.\n- Recipients are drawn only from contacts who have recorded consent (see CRM).\n- Every message waits for human approval; nothing is ever sent automatically.\n\n#### Analytics\n- Captures page views and on-site searches across anything you want to measure.\n- CMS pages are tracked automatically; for any other page, you add a small tracking snippet.\n- Conversion funnels and per-visitor journeys — once someone is identified, their visits link to a CRM contact, including the anonymous ones from before.\n- Breakdowns by traffic source, referrer, and country, plus \"what to write next\" signals (popular topics, engagement, and searches that came back empty).\n\n## Automation\n\n#### Conversation loop\nAn in-process, per-org agent runner answers live conversations on every channel (chat widget, email, SMS, voice) against the LLM provider you configure — drafting and sending replies, and handing off to a human when needed.\n\n#### Curator loop\nThe in-process agent runner also works a durable background job queue: scheduled KB curation, CRM hygiene, contact extraction, stale-content review, and outreach drafts, with retry and dead-letter handling.\n\n#### Playbooks & skills\nPackaged markdown procedures (`skill://module/`) for multi-step, cross-module workflows, surfaced over MCP — followed both by Munin's own runner and by any external AI agent operating on the platform.\n\n## Platform\n\n#### Data portability\nSymmetric `*_export` / `*_import` MCP tools (and `/v1//export|import` REST endpoints) per module, so an agent can move an org's data between a self-hosted server and the cloud in either direction. See `skill://playbooks/data-migration`.\n\n#### Audit & webhooks\nEvery action is written to an audit log, and webhooks fan those events out to your own endpoints with signed, replayable deliveries.\n\n#### Alerts & feedback\nOperational issues surface as system alerts the agent can list, acknowledge, and resolve. An in-product feedback channel lets you file feature requests and vote on Munin's public roadmap.\n\n#### Auth & access\nSign-in and access control run on BetterAuth, with OAuth 2.1 dynamic-client registration and team invites.\n\n## See it in action\n\n> Lovable builds your frontend. Munin spins up your operations. One prompt, one MCP endpoint — and the agents do the rest.\n\nWatch Lovable build a real website from a single prompt while Munin stands up everything behind it — the CMS the blog reads from, a seeded knowledge base, analytics, and a chat widget that already knows the business. No click-ops, no screens to wire up; the agent does the work, over one MCP endpoint. Then a real customer conversation plays out: answered from the knowledge base, handed off to a human when it matters, and picked back up by the agent to close.\n\n

\n \n \"Watch:\n \n

\n\n## Two ways to run\n\n**Self-host** (this repo): single-tenant, invite-only.\n\n```bash\ngit clone https://github.com/getmunin/munin.git\ncd munin\ncp .env.example .env\ndocker compose up\n```\n\nSecrets left at their `.env.example` placeholders are auto-generated on first boot and persisted in the `munin-data` volume — fine for local self-hosting. For shared or production deployments, set strong `MUNIN_AUTH_SECRET` + `MUNIN_KEY_PEPPER` + `MUNIN_ENCRYPTION_KEY` values (`openssl rand -base64 48`) in `.env` instead.\n\nThe first user to sign up becomes the org admin; subsequent users need an invitation token or an email whose domain is in `MUNIN_ALLOWED_EMAIL_DOMAINS`.\n\n**Hosted** (https://www.getmunin.com): multi-tenant, one signup per org.\n\n## Try it locally\n\nAfter `docker compose up`, the backend listens on `:3001` and the dashboard on `:3000`.\n\n1. Open `http://localhost:3000` and register the first user — they become the singleton org admin.\n2. In the dashboard, go to **Settings → API keys** and mint an admin key (`mn_admin_…`). Shown once; treat like a password.\n3. Poke at the API and tools:\n\n```sh\n# REST control plane — direct, no OAuth\ncurl -s http://localhost:3001/v1/kb/spaces \\\n -H \"Authorization: Bearer mn_admin_...\" | jq\n\n# MCP tool browser (recommended for poking at tools/skills)\nnpx @modelcontextprotocol/inspector\n# In its UI: URL = http://localhost:3001/mcp, Auth = Bearer mn_admin_...\n\n# Raw curl over Streamable HTTP — useful for sanity-checking the transport\ncurl -N -X POST http://localhost:3001/mcp \\\n -H \"Authorization: Bearer mn_admin_...\" \\\n -H \"Content-Type: application/json\" \\\n -H \"Accept: application/json, text/event-stream\" \\\n -d '{\"jsonrpc\":\"2.0\",\"id\":1,\"method\":\"tools/list\"}'\n```\n\nThe OpenAPI spec for the REST control plane is at `packages/backend-core/openapi.json`. To wire an MCP client like Claude or Cursor, see [Connect your AI agent](#connect-your-ai-agent) below.\n\n## Connect your AI agent\n\nOnce you've signed up (hosted) or run `docker compose up` (self-host), point your MCP client at the URL — `http://localhost:3001/mcp` for self-host, or `https://mcp.getmunin.com` for hosted.\n\n**Claude Code (CLI):**\n\n```sh\nclaude mcp add munin http://localhost:3001/mcp\n```\n\n**Claude Desktop** — add to your MCP config:\n\n```json\n{\n \"mcpServers\": {\n \"munin\": {\n \"url\": \"http://localhost:3001/mcp\"\n }\n }\n}\n```\n\nThe first call triggers an OAuth consent screen in your browser, then your agent has the full tool surface — Knowledge Base, Conversations, CRM, CMS, Outreach, Analytics.\n\n## Two trust contexts, one MCP endpoint\n\nThe same `/mcp` endpoint serves two distinct callers, audience-aware:\n\n- **Admin agents** (Claude Desktop, Cursor, internal automation) — OAuth-authorized by you. Full tool surface, scope-gated per `kb:*`, `conv:*`, `crm:*`, `cms:*`, `outreach:*`, `analytics:*`.\n- **End-user agents** (your voice AI, web chatbot, mobile app helper) — short-lived delegated tokens minted server-side from your backend, scoped to one of your end-users. Only self-service tools (read your own contact, send a message in your own conversation).\n\nSee `packages/backend-core/src/control/delegated-token.controller.ts` for the token-mint API. The `@getmunin/sdk` Node client wraps it.\n\n## Stack\n\n| Layer | Tech |\n|---|---|\n| Language & runtime | TypeScript, Node 24 LTS |\n| Monorepo | Turborepo, pnpm |\n| Backend | NestJS |\n| Frontend | Next.js |\n| Data | Postgres + pgvector, Drizzle |\n| Protocol & auth | MCP Streamable HTTP, BetterAuth + OAuth 2.1 |\n\n## Documentation\n\nDeveloper docs live at **[getmunin.com/docs](https://www.getmunin.com/en/docs/)** — guides, the REST API reference, the full MCP tool list, and the skill library.\n\n## Contributing\n\nContributions are welcome. `pnpm install`, then `docker compose up` (or `pnpm dev`) gives you a full stack on `:3001` (backend) and `:3000` (dashboard). Branch from `main` as `/` (e.g. `feat/website-import-reconcile`), keep PRs focused, and make sure CI (lint, typecheck, test, build) passes.\n\nSee [CONTRIBUTING.md](./CONTRIBUTING.md) for setup, commit conventions, and PR guidelines.\n\n## Security\n\nFound a vulnerability? Please **don't** open a public issue — email **security@getmunin.com** instead. See [SECURITY.md](./SECURITY.md) for scope and our response timeline.\n\n## License\n\nMIT. See [LICENSE](./LICENSE).\n\nBundled third-party dependencies retain their own licenses — see [THIRD_PARTY_LICENSES.md](./THIRD_PARTY_LICENSES.md) (generated by `pnpm licenses:generate`, verified in CI).\n", "bytes": 12546, "sha": "e12b7ab55105f5f04795039c04841bda98c6654b370d97b119195c9a5b83d799", "repo_slug": "getmunin/munin", "fonte": "repo", "truncated": false, "api": "https://agentalog.com/api/listings/mcp_com_getmunin_munin_8f6b7b85/readme" }