Models and media backends

OpenClaw Providers

Choose model, speech, image, search, and local OpenAI-compatible backends with cost, quality, tool-calling, and fallback behavior in view.

CoClaw is community-run and independent from the official OpenClaw project. Use official OpenClaw docs for canonical command, schema, and release details.

Official docs All CoClaw hubs

Hosted frontier APIs

Best when quality and tool reliability matter most. Budget for usage and keep auth profiles auditable.

Read next

OpenAI-compatible relays

Useful for LiteLLM, vLLM, SGLang, Ollama, and gateway services, but tool, streaming, and reasoning behavior can drift.

Read next

Local/offline paths

Good for privacy and fallback posture. Validate context length, tool calling, and latency before relying on unattended automation.

Read next

CoClaw reading path

Community pages are intentionally narrower than upstream docs. Use them to choose a path, then return to official references for exact commands and schema fields.

Compatibility matrix

CoClaw matrix for relays, local servers, tools, reasoning, and streaming.

Open

Cost and routing strategy

Community notes on fallback, budget, and model selection tradeoffs.

Open

Provider config generator

Build a starter config for built-in or custom provider paths.

Open

Other current OpenClaw surfaces

These CoClaw hubs are first-pass orientation pages, not exhaustive docs.

CoClaw treats OpenClaw v2026.6.1, published June 3, 2026, as the latest stable baseline for this first-pass map. Main branch and beta builds can move faster and should not be presented as stable.

Platforms Start here when the question is macOS app, Windows, Linux, VPS, Raspberry Pi, iOS, Android, or cloud hosting. Nodes Use this hub when a workflow depends on a paired device, node permissions, Canvas, media capture, or remote browser control. Tools Use this hub when OpenClaw can chat but cannot act, or when an operator is deciding which actions should be available. Plugins Use this hub when a capability lives outside the core Gateway or when an operator is evaluating ClawHub packages. Automation Use this hub when OpenClaw is expected to initiate or continue work outside a single live chat reply.