OpenClaw Provider Compatibility Checker
Check your AI provider's compatibility with OpenClaw. Validate transport layers, cache semantics, endpoint detection, and streaming support — then generate a safe config snippet. Based on 8+ provider PRs merged on April 3, 2026.
What the Compatibility Checker Validates
Every check maps to real OpenClaw PRs merged on April 3, 2026 — transport fixes, cache alignment, endpoint centralization, and more.
Transport Compat Validation
Check if your provider's SSE/streaming transport layer is fully supported, partially compatible, or requires the compat shim — based on Mistral (PR #60405) and Moonshot (PR #60411) fixes.
Cache Semantics Check
Verify cache support levels: native caching (Anthropic), passthrough via custom APIs (PR #60359), scoped family semantics (PR #60370), and TTL alignment (PR #60375).
Endpoint Detection Status
See if your provider uses a native endpoint handler or the centralized compat endpoint detection layer — now unified in PR #60399 and #60341.
Safe maxTokens Guardrails
Get the correct maxTokens value for each model — avoid the common 422 error from setting maxTokens to the context window size, especially on Mistral models.
Replay Policy Awareness
Check whether your provider uses plugin-managed replay policies (PR #60452) or built-in handling, and get the right plugin config if needed.
Config Snippet Generator
Generate a ready-to-use OpenClaw config snippet with safe maxTokens, correct streaming settings, cache config, and plugin requirements for your chosen provider and model.
Check Provider Compatibility
Select your provider and model to see the full compatibility breakdown, warnings, and a generated config snippet.
Select Provider & Model
Context window
1.0M
Safe maxTokens
32k
Compat score
A (100%)
Compatibility Matrix
Based on provider-specific PRs merged on April 3, 2026 — centralized compat detection, transport fixes, cache semantics, and role support.
Native Endpoint Detection
Provider has a dedicated endpoint handler (PR #60341)
Transport Compatibility
SSE/streaming transport layer support (PRs #60405, #60411)
Cache Support
Prompt caching capabilities (PRs #60359, #60370, #60375)
Cache TTL Semantics
TTL alignment with provider expectations (PR #60375)
Developer Role
Support for developer-role in responses (PR #60385)
Replay Policy
Conversation replay policy management (PR #60452)
Legacy Streaming Aliases
Requires legacy streaming config aliases (PR #60358)
Claude Opus 4.6 — Compatibility Notes
- Cache retention passthrough via custom APIs (PR #60359)
- Scoped cache semantics for Anthropic-family (PR #60370)
- Cache TTL aligned with provider semantics (PR #60375)
Anthropic — Warnings & Breaking Changes
- Cache retention now passes through custom APIs — ensure your proxy preserves the cache-control header (PR #60359)
- Auth permanent failures now use shorter backoff — failed keys are detected faster (PR #60404)
Generated Config Snippet
Ready-to-use config for Claude Opus 4.6 via Anthropic with safe maxTokens and recommended settings.
{
"agents": [
{
"name": "Claude Opus 4.6 Agent",
"provider": "anthropic",
"model": "claude-opus-4-6",
"maxTokens": 32000,
"cache": {
"enabled": true,
"ttl": "auto"
}
}
],
"plugins": {
"entries": [
"@openclaw/plugin-replay"
]
}
}Deploy OpenClaw with DeployClaw
All provider compatibility handled automatically — transport, caching, streaming, and auth.
Frequently Asked Questions
What does this tool check?
It validates 7 compatibility dimensions for each AI provider: native endpoint detection, transport compatibility (SSE streaming), cache support level, cache TTL semantics, developer-role support, replay policy type, and legacy streaming alias requirements. Each check maps to specific OpenClaw PRs merged on April 3, 2026.
Why do some providers show 'Partial' transport compat?
Providers like Mistral and Moonshot use non-standard SSE streaming formats. OpenClaw's transport compat layer (PRs #60405, #60411) translates these to the expected format, but edge cases may still require the compat shim. 'Full' means the provider's streaming works natively with OpenClaw.
What happens if I set maxTokens to the context window size?
Some providers (especially Mistral) interpret maxTokens as the output budget, not a cap on context. Setting it to 262,144 on Mistral Large makes the API try to generate 262k output tokens, which immediately fails with a 422 error. Always use the 'Safe maxTokens' value shown in this tool.
What is the centralized compat endpoint?
PR #60399 consolidated compatibility endpoint detection into a single module. Providers without native endpoint handlers (like DeepSeek, xAI) are routed through this centralized layer, which normalizes their API differences. Providers with native detection (Anthropic, OpenAI, Mistral, Moonshot, Google) use their own optimized handlers.
What changed with replay policies?
PR #60452 moved bundled replay policy ownership into plugins. Providers that previously had built-in replay handling (Anthropic, OpenAI) now manage it through the @openclaw/plugin-replay plugin. This gives you more control over replay behavior and makes it configurable per-agent.
Is my configuration data sent anywhere?
No. The compatibility checker runs entirely in your browser. Provider data and compatibility flags are embedded in the page. The generated config snippet is assembled client-side. Nothing is transmitted to any server.