OpenClaw
A self-hosted personal AI agent that stays available across messaging channels, tools, skills, memory, and local workspaces.
By OpenClaw contributors · reviewed 2026-10-21
What OpenClaw does
OpenClaw is an always-on personal agent platform designed to connect model providers with messaging channels, local files, browser actions, scheduled work, memory, and installable skills. It can use hosted or local model endpoints, but broad tool access makes isolation, permissions, secrets, and backup policy essential parts of deployment.
Pricing model
Free and open source; model and integration costs may apply
Position in the stack
Where OpenClaw fits
OpenClaw sits at the personal-agent application layer. It coordinates channels, tools, skills, memory, scheduled work, and model providers; it does not replace the runtime or API that performs inference.
Typical deployment flow
A practical sequence for evaluating OpenClaw before making it part of a permanent stack.
- STEP 1
Choose the trust boundary
Decide which machine, accounts, folders, and channels the agent may access.
- STEP 2
Connect a model path
Configure a hosted provider or local runtime whose model can use the required tools.
- STEP 3
Add one workflow at a time
Enable skills gradually, preserving logs, approvals, backups, and a recovery path.
Best for
- An always-on personal agent across chat channels
- Scheduled and event-driven personal automation
- Extending one agent with reusable skills
Not the right layer for
- Untrusted public access without strong isolation
- Teams needing mature centralized governance by default
Capabilities
Capabilities refer to the tool's application layer. Hardware and model support can still depend on a connected inference engine.
Deployment and compatibility
- Deployment
- Desktop / localSelf-hosted
- Platforms
- macOS · Windows · Linux
- Hardware backends
- Uses connected model provider
- Install methods
- Installer · npm · Docker · Build from source
Models and integrations
- Model formats
- Uses connected local or hosted model provider
- Common integrations
- Messaging channelsSkillsBrowser toolsLocal workspaceOllamaOpenAI-compatible APIs
Strengths
- Persistent personal-agent workflow
- Large skill and channel surface
- Can connect to local and hosted models
Limitations
- Large permission and security surface
- Reliable operation requires backups, secrets management, and careful upgrades
Planning checklist
Before you choose OpenClaw
Answer these questions with the exact models, hardware, users, and data you expect to operate.
- Which credentials and files can the agent reach?
- Which actions require approval?
- Does the model reliably call the required tools?
- How will persistent state be backed up?
OpenClaw FAQ
Does OpenClaw run models itself?
No. It orchestrates a connected hosted provider or local inference server.
Why is isolation important?
Browser, shell, file, and credential access can turn a bad instruction into a real system change.
Is it only for chat?
No. Its main value is persistent, tool-using automation across channels and schedules.
Official sources
Use these links to confirm current compatibility and installation requirements.
Related tools
A desktop and self-hosted workspace for document chat, RAG, and AI agents.
A node-based workflow engine and interface for local generative image and media models.
An open-source, plugin-first agent harness with local Web UI, tools, sessions, approvals, and extensible model adapters.
Comparisons featuring OpenClaw
Use a reviewed comparison when the choice is between two adjacent tools.
Execution evidence
Known working recipes using OpenClaw
Recipes connect hardware, a model artifact, tools, settings, verification, and a reportable result.
No verified recipe is linked to this record yet.
Compatibility estimates remain available in the planner. A recipe appears here only after its exact stack and verification protocol are documented.