Bug type

Behavior bug (incorrect output/state without crash)

Beta release blocker

No

Summary

OpenClaw's LM Studio provider fails due to two critical issues:

Environment variables ($LMSTUDIO_API_KEY) not expanded in API calls
API endpoint incompatibility (uses /v1/chat/completions but LM Studio's OpenAI-compatible API returns HTTP 200 with errors)

Steps to reproduce

1. Configure LM Studio provider in openclaw.json with apiKey: "$LMSTUDIO_API_KEY"
2. Set environment variable LMSTUDIO_API_KEY with valid token
3. Attempt to use any LM Studio model
4. Observe 401 error with literal string $LMSTUDIO_API_KEY

Expected behavior

• Environment variables should be expanded before sending API requests
• Should detect and use LM Studio's working endpoints (/api/v1/chat for native API or /v1/completions for OpenAI-compatible)

Actual behavior

• $LMSTUDIO_API_KEY sent literally, causing 401 errors
• Attempts to use /v1/chat/completions which LM Studio returns HTTP 200 with error: "Unexpected endpoint or method

OpenClaw version

2026.5.7

Operating system

Linux (containerized)

Install method

docker

Model

google/gemma-4-26b-a4b

Provider / routing chain

openclaw -> LM Studio

Additional provider/model setup details

Additional Context:

• LM Studio has functional native API at /api/v1/chat (different format)
• LM Studio's OpenAI-compatible API at /v1/chat/completions is buggy (returns 200 with errors)
• LM Studio's old OpenAI completions API at /v1/completions works but uses deprecated format
• Other providers (OpenRouter, Gemini) properly expand environment variables

Proposed Solutions:

• Fix environment variable expansion in LM Studio provider
• Add support for LM Studio's native API (/api/v1/chat)
• Implement endpoint fallback logic for LM Studio's inconsistent API

Logs, screenshots, and evidence

Impact and severity

• Priority: High - prevents users from using local LM Studio inference with OpenClaw

Additional information

Related Links:

• https://docs.openclaw.ai/ (OpenClaw documentation)
• https://lmstudio.ai/docs/developer/core/authentication (LM Studio API docs)