OwnPilot Personal AI Assistant
A self-hosted personal AI assistant that brings models, tools, and automation into your own environment.
- Source repo
- ownpilot/OwnPilot
- Stars
- ★ 429
- Last updated
- 1mo ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 61/100 · Some gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformOpenAI API · Claude API
- Cost
- Free software; you pay for model usage
- Setup effort
- High · needs real infrastructure
- You'll need
- Typical use
- Individuals who want tasks, notes, calendars, contacts, and long-term memory in a self-hosted assistant.
- Not a fit if
- Users who require hosted SaaS with no service to maintain
- Teams unable to run PostgreSQL or a local service
- Source review
- 61/100 · Some gaps
What does this agent do, and when should you use it?
OwnPilot is a privacy-focused, self-hosted personal AI assistant platform built from a Gateway API service, React web UI, CLI, and Core tool framework. It calls models through native providers including OpenAI, Anthropic, Google, and Zhipu, supports OpenAI-compatible endpoints, and can discover local services such as Ollama, LM Studio, LocalAI, and vLLM. Standard agents, Soul Agents, Claw Agents, and crews can use tools, run on schedules, or respond to events; chat supports SSE streaming and WebSocket updates. The platform also includes Telegram and WhatsApp channels, MCP integration, browser automation, voice features, and MQTT edge-device control. Deployment requires a running service and PostgreSQL; Docker Compose is the documented quick-start path, with data and configuration stored in the self-hosted environment.
Users submit requests through web chat, Telegram, WhatsApp, or the REST API; the Gateway MessageBus pipeline handles audit, persistence, context injection, and agent execution. An agent can find tools with search_tools, inspect them with get_tool_help, then run them with use_tool or batch_use_tool—for example, to read or write files, manage tasks and notes, search the web, run code, or call connected MCP tools. Claw Agents can run single-shot, continuous, interval, or event-driven tasks in isolated workspaces, using directive files to retain context across cycles. Users can also configure triggers, heartbeats, plans, and workflows, then review approvals, activity logs, and versioned artifacts.
- Individuals who want tasks, notes, calendars, contacts, and long-term memory in a self-hosted assistant.
- Individuals or small teams that need agents to check information on a schedule, follow checklists, and log activity.
- Users who want to interact with a self-hosted assistant through Telegram or WhatsApp.
- Developers who need MCP servers, browser automation, or local CLI tools for multi-step work.
- IoT developers who need MQTT sensor and actuator connections, device status checks, or command delivery from an assistant.
How do you install or deploy this agent?
The quick-start path uses Docker Compose to start OwnPilot and PostgreSQL:
git clone https://github.com/ownpilot/ownpilot.git
cd ownpilot
docker compose --profile postgres up -dThen open http://localhost:8080. Source installation is also documented and requires Node.js >=22.0.0, pnpm >=10.0.0, and PostgreSQL 16+; an interactive setup.sh is provided.
How do you use this agent?
On first use, configure an AI provider API key in the web Config Center; keys and settings are stored in PostgreSQL. You can also configure a key and start the service with the CLI:
ownpilot setup
ownpilot config set openai-api-key sk-...
ownpilot startOpen http://localhost:8080 for the web UI. In development, pnpm dev starts the Gateway and Vite UI at http://localhost:8199. Discover and run tools through the chat tool proxy, or add an external MCP service under Settings → MCP Servers.
What are this agent's strengths and limitations?
- Offers multi-provider routing, local model services, and arbitrary OpenAI-compatible endpoints.
- Includes 250+ built-in tools, an MCP client and server, and custom-tool and extension mechanisms.
- Documents security controls including permission policies, execution approval, sandboxing, PII redaction, and tamper-evident audit.
- Supports scheduled or event-driven Soul Agents, Claw Agents, crews, and workflows.
- Deployment depends on Node.js, pnpm, and PostgreSQL; source setup involves several steps.
- Users configure their own model API keys, and hosted model use may incur separate costs.
- The README does not document a hosted SaaS option; users deploy and maintain the service and database.
- The broad feature set means channels, MCP, MQTT, and automation may each require additional service or credential configuration.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| OwnPilot Personal AI Assistant This agent | 61 · Some gaps | Self-hosted serviceFree + model costs | ★ 429 | 1mo ago | TypeScript | OpenAI API · Claude API |
| GAIA — Personal AI Assistant | 73 · Some gaps | Hosted serviceFreemium | ★ 324 | today | Python | — |
| LinkedIn MCP Server | 79 · Good | MCP serverFree | ★ 4.1k | today | Python | — |
| Mnemoverse Memory | 77 · Some gaps | MCP serverFreemium | ★ 28 | today | TypeScript | ChatGPT · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The materials describe risk-based tool policies, approval flows, skill permission review, sandboxing, audit logs, and encryption, and state that defaults target local single-user deployment. They also disclose limits such as name-based permission gates and filesystem checks that may fail open. Scores are moderate because controls are concrete but their coverage and defaults remain limited; general recovery mechanisms are unclear, and attribution is limited to repository and author metadata while publisher identity is unverified.
The README, package manifest, and CI workflow describe build, typecheck, lint, test, and database startup checks, alongside runtime and dependency requirements. Deductions reflect residual issues documented in the security audit, including fail-open permission behavior and heuristic rules; the supplied materials do not describe failure handling in equal detail for the wide range of advertised features.
The materials cover personal-assistant and self-hosted use, multiple providers, autonomy levels, scheduled and event-driven triggers, containers, and local models, with Docker, CLI, and source installation paths. Deductions reflect stated boundaries, including the local single-user posture for no-password deployment and dependencies on external services or local configuration.
The README provides substantial navigation, architecture, project structure, setup, and configuration guidance, and the MIT license is explicit. Security policy, limitation notes, and a changelog entry point are also present. Deductions reflect that the supplied material does not include a complete FAQ or changelog, or a clearly identified maintenance owner; version information exists, but maintenance routes mainly through the repository and security advisories.
The platform combines personal data, tool orchestration, autonomous agents, and multiple interaction channels in a self-hosted product, with useful delivery paths such as chat, notifications, and artifacts. Deductions reflect the broad feature scope and operational requirements, including PostgreSQL, Node/pnpm or Docker, and potentially model APIs and messaging services; costs and benefits are not quantified.
Some major README claims are corroborated by the package manifest, CI workflow, security policy, and repository audit/remediation records. The audit distinguishes findings, fixes, and residual limitations. Deductions reflect that many scale and capability figures remain README assertions, and the audit and remediation records are repository-authored rather than independent verification.
- The security audit documents name-based tool permission gates, filesystem path checks that may fail open, and heuristic detection of destructive shell actions; review policies against the tools and network exposure in your deployment.
- The security policy scopes no-password mode to local single-user use; configure authentication, a UI password, trusted-proxy settings, and TLS-related options before exposing the gateway.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
What's Changed * fix(security): authenticate /api/v2/* and apply security-check hardening by @ersinkoc in https://github.com/ownpilot/OwnPilot/pull/105 * fix(ui): gate LLM-controlled FileWidget download URL against javascript: URIs by @ersinkoc in https://github.com/ownpilot/OwnPilot/pull/106 * chore(release): v0.8.3 by @ersinkoc in https://github.com/ownpilot/OwnPilot/pull/107 Full Changelog: https://github.com/ownpilot/OwnPilot/compare/v0.8.2...v0.8.3View on GitHub ↗ -
Release notes excerpt
Full Changelog: https://github.com/ownpilot/OwnPilot/compare/v0.8.1...v0.8.2View on GitHub ↗ -
Release notes excerpt
Full Changelog: https://github.com/ownpilot/OwnPilot/compare/v0.7.4...v0.8.1View on GitHub ↗