@yadimon/claude-code-to-llm-server

OpenAI-compatible Responses server on top of @yadimon/claude-code-to-llm.

Install

npm install -g @yadimon/claude-code-to-llm-server

Or run it without installing globally:

npx @yadimon/claude-code-to-llm-server

Requirements:

• Node.js >=20
• installed claude CLI in PATH or CLAUDE_CODE_TO_LLM_CLI_PATH
• valid Claude Code login on the machine

Default auth bundle:

• ~/.claude.json
• ~/.claude/.credentials.json

Endpoints

• POST /v1/responses
• GET /v1/models
• GET /healthz

Start

npx @yadimon/claude-code-to-llm-server

Then call:

curl http://127.0.0.1:3000/healthz
curl http://127.0.0.1:3000/v1/models

Example response request:

curl http://127.0.0.1:3000/v1/responses \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "input": "Say hello in one short sentence."
  }'

Streaming example:

curl http://127.0.0.1:3000/v1/responses \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "stream": true,
    "input": "Count from 1 to 3."
  }'

Local development commands:

npm run start --workspace @yadimon/claude-code-to-llm-server
npm run start:mock --workspace @yadimon/claude-code-to-llm-server

Authentication

If you set CLAUDE_CODE_TO_LLM_SERVER_API_KEY, only POST /v1/responses requires a bearer token. GET /healthz and GET /v1/models stay public.

Example:

curl http://127.0.0.1:3000/v1/responses \
  -H "Authorization: Bearer your-token" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "input": "Hello"
  }'

Runtime Configuration

| Variable | Default | Description | |---|---|---| | CLAUDE_CODE_TO_LLM_SERVER_HOST | 127.0.0.1 | HTTP bind host. | | CLAUDE_CODE_TO_LLM_SERVER_PORT | 3000 | HTTP bind port. | | CLAUDE_CODE_TO_LLM_SERVER_DEFAULT_MODEL | claude-sonnet-4-6 | Fallback model when the request omits model. | | CLAUDE_CODE_TO_LLM_SERVER_MODELS | default model | Comma-separated allowlist of accepted models. | | CLAUDE_CODE_TO_LLM_SERVER_API_KEY | - | Bearer token accepted for POST /v1/responses. | | CLAUDE_CODE_TO_LLM_SERVER_MOCK_MODE | - | Enables the mock runner for local testing. | | CLAUDE_CODE_TO_LLM_SERVER_MOCK_RESPONSE | mock response | Mock response text returned by the mock runner. | | CLAUDE_CODE_TO_LLM_AUTH_PATH | ~/.claude.json | Path to the Claude Code session file. | | CLAUDE_CODE_TO_LLM_CREDENTIALS_PATH | ~/.claude/.credentials.json | Path to the Claude Code credentials file. | | CLAUDE_CODE_TO_LLM_SETTINGS_PATH | ~/.claude/settings.json when present | Optional settings file copied into the temporary Claude home. | | CLAUDE_CODE_TO_LLM_CLI_PATH | claude | Path to the Claude Code CLI binary. | | CLAUDE_CODE_TO_LLM_CONFIG_HOME | temp dir | Temporary Claude Code home directory for a run. | | CLAUDE_CODE_TO_LLM_WORKSPACE | temp dir | Workspace directory used for execution. | | CLAUDE_CODE_TO_LLM_REASONING_EFFORT | low | Default reasoning effort passed to the core runner. |

Behavior Notes

• GET /healthz and GET /v1/models stay public even when bearer auth is configured.
• POST /v1/responses validates requested models against CLAUDE_CODE_TO_LLM_SERVER_MODELS.
• max_output_tokens and reasoning.effort are forwarded to the core runner.
• Unsupported request fields such as tools, tool_choice, or input_image return 400.
• The server owns prompt adaptation for instructions and multi-message dialog input before calling the raw core runner.
• --search / webSearch: true is a server-wide policy: when enabled it allows the underlying claude CLI to use WebSearch for every request. There is no per-request opt-in (tools is rejected). Web-search usage is captured by the core runner but not surfaced in the OpenAI usage block.

Docker

Build from the repository root:

docker build -f packages/claude-code-to-llm-server/Dockerfile .
docker run -p 3000:3000 \
  -v ~/.claude.json:/root/.claude.json:ro \
  -v ~/.claude/.credentials.json:/root/.claude/.credentials.json:ro \
  claude-code-to-llm-server

Development

npm run build --workspace @yadimon/claude-code-to-llm-server
npm run lint --workspace @yadimon/claude-code-to-llm-server
npm run typecheck --workspace @yadimon/claude-code-to-llm-server