Use LithosAI with Claude Code

Before you begin

Create a key on the API Keys page, confirm the model ID on the Models page, and export the key in the shell where you start Claude Code or its local proxy.

export LITHOSAI_API_KEY="your-key"

Claude Code speaks the Anthropic Messages API, so its proxy needs a different configuration from the one Codex uses. Run it on a second port to use both agents at once.

This is a community compatibility path for non-Claude models, not a configuration supported by Anthropic. LiteLLM maps common Messages API behavior to Chat Completions, but native thinking blocks, prompt caching, token counting, streaming details, and some tool-use flows may behave differently.

Install LiteLLM

LiteLLM bridges the Anthropic Messages API to Chat Completions.

pip install 'litellm[proxy]'

Install a current release. LiteLLM 1.82.7 and 1.82.8 were briefly replaced on PyPI in March 2026 by a credential stealer.

Configure the proxy

Give each model the hosted_vllm/ prefix. It translates /v1/messages requests into Chat Completions requests for LithosAI.

claude-code.litellm.yaml

model_list:
  - model_name: kimi-k3
    litellm_params:
      model: hosted_vllm/moonshotai/Kimi-K3
      api_base: https://api.lithosai.cloud/v1
      api_key: os.environ/LITHOSAI_API_KEY
  - model_name: kimi-k3-fast
    litellm_params:
      model: hosted_vllm/moonshotai/Kimi-K3-fast
      api_base: https://api.lithosai.cloud/v1
      api_key: os.environ/LITHOSAI_API_KEY

litellm_settings:
  drop_params: true

general_settings:
  master_key: sk-claude-proxy

Do not reuse the Codex configuration here. An openai/ entry with use_chat_completions_api answers /v1/messages with 200 OK and empty content, so Claude Code appears to hang rather than reporting an error.

Start the proxy

litellm --config claude-code.litellm.yaml --host 127.0.0.1 --port 4001

Leave it running, and confirm it is up. The proxy holds your LithosAI key, so keep it bound to 127.0.0.1 and do not expose it to a network.

curl http://127.0.0.1:4001/health/liveliness

Point Claude Code at it

export ANTHROPIC_BASE_URL="http://127.0.0.1:4001"
export ANTHROPIC_AUTH_TOKEN="sk-claude-proxy"
export ANTHROPIC_MODEL="kimi-k3"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="kimi-k3-fast"
claude

ANTHROPIC_DEFAULT_HAIKU_MODEL handles background work such as session titles. Without it, those requests fail against a model the proxy does not serve.

Setting ANTHROPIC_AUTH_TOKEN overrides a Claude.ai login for that shell. Export these in a dedicated shell or profile to keep an existing subscription usable elsewhere.

Claude Code does not recognize LithosAI model names, so it assumes a 200k-token context window and compacts early. Set the real window to avoid that.

export CLAUDE_CODE_MAX_CONTEXT_TOKENS=262144

Check the Models console page for each model’s context window.

Troubleshooting

Symptom	Cause
Empty replies, or the session appears to hang	The model entry uses `openai/` with `use_chat_completions_api`. Claude Code needs `hosted_vllm/`.
Background requests fail while chat works	`ANTHROPIC_DEFAULT_HAIKU_MODEL` is unset or names a model the proxy does not serve.
`400` with `No connected db.`	`ANTHROPIC_AUTH_TOKEN` does not match the proxy’s `master_key`. LiteLLM reports the lookup failure, not an auth error.

Issues common to every agent are listed under Common issues.

Getting Started

Inference APIs

Account

Use LithosAI with Claude Code

Before you begin

Troubleshooting

​Before you begin

​Troubleshooting

Before you begin

Troubleshooting