Manor AI v0.1.0 is the first public source-available release of the self-hosted AI workspace runtime.
Highlights
- One workspace for chat, agents, goals, tasks, documents, knowledge, reports, workflows, and integrations.
- BYOK model configuration so provider credentials remain in your deployment.
- Durable agent work with task status, evidence, comments, approval checkpoints, and audit-friendly history.
- Human-in-the-loop governance with scoped tools, permissions, approval rules, and runtime signals.
- PostgreSQL + pgvector knowledge retrieval, MinIO object storage, Redis, Celery workers, and an isolated sandbox service.
- Extensible skills, tools, webhooks, OAuth/Nango connectors, API keys, and MCP server surfaces.
- Docker Compose deployment path plus public quickstart, configuration, security, backup, and upgrade documentation.
Quick start
git clone https://github.com/manor-os/manor-ai.git
cd manor-ai
cp .env.example .env
docker compose up --build -dOpen http://localhost:18080, sign in with the local demo account shown in the README, and configure your model provider keys in Settings.
Before production use
- Replace every default secret in
.env. - Set
MANOR_DEMO_ACCOUNT_ENABLED=false. - Put the application behind HTTPS and keep PostgreSQL, Redis, MinIO, and sandbox services on private networks.
- Back up PostgreSQL and object storage together.
- Start with narrow agent tool scopes and require human approval for sensitive or externally visible actions.
Known limitations
- This is the first public release; operator feedback and bug reports are welcome.
- The public tree focuses on self-hosting and local evaluation. Managed hosting and private cloud operations are separate.
- Model provider keys must be configured by the operator before AI workflows can run.
- The project is source-available under the Manor Sustainable Use License 1.0. Resale, white-labeling, and competing hosted services are restricted.
Resources
Full changelog: https://github.com/manor-os/manor-ai/commits/v0.1.0