"provider rejected the request schema or tool payload"
On OpenClaw 2026.5.3+ routing DeepSeek V4 Pro through OpenRouter, the gateway internally maps reasoning_effort to 'max', a value OpenRouter doesn't accept, so it rejects the request with a 400. The cleanest fix is to use the native DeepSeek provider at api.deepseek.com. Alternatively set thinkingDefault to "none" to bypass reasoning_effort, or downgrade to 2026.4.10.
Likely cause
On OpenClaw 2026.5.3+ routing DeepSeek through OpenRouter, the gateway sends an invalid reasoning_effort value ('max') that OpenRouter rejects with a 400, surfaced as 'provider rejected the request schema or tool payload'.
The fix
- 1 Switch to the native DeepSeek provider (api: deepseek, baseUrl: https://api.deepseek.com) instead of OpenRouter.
- 2 Or set thinkingDefault to "none" in your config to bypass reasoning_effort.
- 3 Or downgrade to OpenClaw 2026.4.10.
"thinkingDefault": "none"Hit a different error?
Paste any agent error and get the cause and fix in seconds.
Frequently asked questions
My config says 'high', which is valid. Why is it rejected?
The gateway silently maps your value to 'max' internally before sending it to OpenRouter, and OpenRouter only accepts xhigh, high, medium, low, minimal, or none. The error blames your config even though the gateway sent the bad value.
Will switching to the direct DeepSeek API fix it?
Yes. The native DeepSeek provider doesn't hit the reasoning_effort bug and also handles reasoning_content replay, so multi-turn tool use works out of the box.
It worked on an older OpenClaw version. Can I just downgrade?
Yes. The regression landed in 2026.5.3-1; downgrading to 2026.4.10 restores the working behavior until the OpenRouter path is patched.
Stop firefighting agent errors
Decoding errors one at a time is the manual version of what BetterClaw automates. Run your OpenClaw agents hosted with managed models, retries and config validation built in.
$19/month per agent · BYOK · 7-day money-back guarantee