Manor AI logo

Manor AI

Self-hosted AI workspace for governed agents, tasks, tools, and knowledge.

Run agents against team knowledge and scoped tools, require human approval before sensitive actions, and keep operational data inside infrastructure you control.

⭐ Star Manor AI · 🚀 Run with Docker Compose · 💬 Open an issue

5-minute Quickstart · Docs · GitHub · Discord · Twitter · Website

Why Manor AI

Most AI products stop at chat. Real teams need an operating layer around AI:
shared context, durable tasks, scoped tools, approvals, audit trails, and a way
to keep sensitive data inside infrastructure they control.

Manor AI is built for that layer: the product surface where model choice,
workspace memory, tasks, permissions, and human review all meet.

Own the runtime, not just the chat box

Manor AI brings the core surfaces of an AI workspace into one runtime: chat,
documents, agents, goals, workflows, reports, integrations, settings, and the
infrastructure around them. Self-host the app, database, files, agent runtime,
and integration surfaces; bring your own model keys instead of handing your
workspace to a hosted black box. Model provider keys are configured in your
deployment, and operational data stays inside your infrastructure.

Turn prompts into accountable work

Agents operate through goals, tasks, workspace context, tool permissions, and
human approval gates. Important automation leaves durable status, evidence, and
review points instead of disappearing into a one-off chat transcript. Goals
break work into linked execution steps that teams can inspect and improve.

Make governance part of the product

Approval gates, scoped tools, audit logs, workspace permissions, and runtime
signals make automation inspectable before it touches critical workflows. Rules
are visible to operators and mapped to enforcement patterns in the runtime.

How Manor AI works

Manor AI is organized around a workspace. A workspace is the boundary where
people, agents, documents, tasks, goals, tools, credentials, and governance
rules come together.

Browser workspace
  |
  v
React web app
  |
  v
FastAPI control plane  <----> PostgreSQL + pgvector
  |                       |   Redis
  |                       |   MinIO / JuiceFS
  v
Worker runtime
  |
  +-- Agent task execution
  +-- Skills and scoped tools
  +-- Sandbox service
  +-- Webhooks / OAuth / Nango integrations
  +-- Human approval checkpoints

Concept	What it means in Manor AI
Workspace	The operating boundary for people, agents, documents, tasks, knowledge, credentials, and audit history.
Agent	A reusable AI worker with instructions, model preferences, tool bindings, skills, and governance rules.
Goal / task	Durable work objects that turn prompts into trackable execution, evidence, comments, approvals, and status.
Knowledge	Uploaded documents and extracted text indexed for workspace retrieval with PostgreSQL and pgvector.
Tools / skills	Bounded capabilities and instruction packages that define what an agent can do and how it should do it.
HITL governance	Approval and deny policies that pause sensitive actions before agents affect external systems or critical data.
Integrations	Webhooks, OAuth/Nango connectors, API keys, and external callbacks for connecting Manor AI to your stack.

Core capabilities

Capability	What you get
AI workspace	Chat, goals, tasks, documents, knowledge, reports, agents, and settings in one self-hosted UI.
Agent runtime	Tool-calling loop, model routing, skills, workspace context, task execution, and evidence logs.
BYOK model access	Provider keys are configured in your deployment; Manor AI does not require a hosted model proxy.
Governance	Human approval gates, scoped tools, workspace permissions, runtime signals, and audit-friendly task history.
Knowledge and files	Document uploads, generated artifacts, pgvector-backed retrieval, MinIO object storage, and entity filesystem support.
Integrations	Webhooks, OAuth provider configuration, optional Nango, API keys, MCP server catalog, and connector surfaces.
Operations	Docker Compose stack, health checks, backup guidance, sandbox isolation, configuration docs, and upgrade notes.
Extensibility	Add skills, tools, integrations, workers, model providers, and API clients without changing the core workspace model.

5-minute Quickstart

git clone https://github.com/manor-os/manor-ai.git && cd manor-ai
cp .env.example .env
docker compose up --build -d

Open http://localhost:18080.

Self-hosted mode seeds a local demo account by default:

[email protected] / manor-demo

After signing in, add your model provider keys in Settings. Manor AI is BYOK in
self-hosted deployments; provider credentials stay in your deployment.

First workflow to try

Use the first run to verify the whole system, not just the login page:

Start the Docker Compose stack and sign in with the demo account.
Add a model provider key in Settings.
Open the seeded workspace and inspect goals, tasks, documents, and rules.
Create or run a task so an agent uses workspace context and tools.
Trigger a governed action and confirm it pauses for human approval.
Review the resulting task status, evidence, comments, and audit trail.

What ships in the OSS self-hosted stack

Area	Included
Workspace app	React + Vite web UI for chat, tasks, agents, knowledge, documents, workflows, reports, and settings.
API runtime	FastAPI service with auth, RBAC, audit logging, OpenAPI docs, and workspace APIs.
Worker runtime	Celery-backed worker for background jobs, agent execution, task runs, and integration callbacks.
Data services	PostgreSQL 16 with pgvector, Redis, MinIO, and optional JuiceFS-backed entity storage.
Agent runtime	Tool-calling loop, skills, scoped tools, HITL approvals, task runners, and goal workflows.
Sandbox	Isolated execution service for code-producing tools and file-generating workflows.
Integrations	Webhooks, OAuth provider configuration, optional Nango, API keys, MCP catalog, and connector surfaces.
Docs and operations	Docusaurus docs, Docker Compose guide, configuration guide, security notes, backup/restore, and upgrade docs.

The public OSS tree is designed for self-hosted deployments and local
evaluation. Managed hosting, private deployment automation, and commercial
cloud operations are separate from this repository. If you are extending Manor
AI, start from the OSS surfaces above: tools, skills, workers, webhooks,
OpenAPI clients, and deployment configuration.

Extend Manor AI

Add skills to teach agents domain-specific workflows and tool usage
boundaries.
Bind tools to agents so each worker has only the capabilities it needs.
Connect systems through webhooks, OAuth provider configuration, Nango, API
keys, and MCP servers.
Use the API through the same FastAPI surface that powers the web app; see
the API reference.
Customize deployment with .env, Docker Compose profiles, storage,
sandbox settings, and provider credentials.

Deploy and operate

Before sharing a deployment with real users:

Replace default secrets in .env.
Use HTTPS for APP_URL and PUBLIC_BASE_URL.
Keep PostgreSQL, Redis, MinIO, and sandbox services on private networks.
Back up PostgreSQL and object storage together.
Treat shell/code execution as sensitive and pair it with HITL governance.
Start with narrow agent tool scopes, then loosen only after workflows are
trusted.

Useful docs:

Contributing

For code changes, local development setup, tests, and style guidance, start
with CONTRIBUTING.md and the
development docs. Keep
changes focused, add tests for behavior changes, and update documentation when
setup or user-facing behavior changes.

Community

Join the Discord.
Follow updates on Twitter / X.
Visit the Manor AI website.
Use GitHub issues for bugs and feature requests.
Report suspected vulnerabilities privately through SECURITY.md.

License

Manor AI is fair-code: source-available, self-hostable, and extensible under the
Sustainable Use License 1.0. Internal business, personal, and non-commercial use
are permitted. White-labeling Manor AI, charging others for access to a hosted
Manor AI service, or otherwise providing the software commercially requires a
separate written agreement with Manor AI. Commercial consulting and support
around a customer's internal deployment are permitted. The Manor AI name and
marks are governed separately by TRADEMARKS.md.

The license text follows
n8n's published Sustainable Use License 1.0.