Use LithosAI with Codex

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 Codex or its local proxy.

export LITHOSAI_API_KEY="your-key"

Codex no longer speaks Chat Completions. wire_api = "chat" is rejected at startup, and responses is the only accepted value, so Codex reaches LithosAI through a proxy that accepts Responses API requests.

This is a compatibility path, not native Responses API support. LiteLLM translates common text and function-tool requests, but Responses-only tools and metadata may not map to Chat Completions. If your workflow depends on a specific Codex tool, verify it before relying on this setup.

Install LiteLLM

LiteLLM bridges the Responses 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

The model below is an example, not the full LithosAI catalog. For another model or tier, replace the ID after openai/ with one from Models. Keep the proxy’s model_name and the Codex model setting consistent.Give the model the openai/ prefix and set use_chat_completions_api, which bridges incoming /v1/responses requests to LithosAI’s /v1/chat/completions.

LiteLLM proxy configuration

codex.litellm.yaml

model_list:
  - model_name: deepseek-v4.1-flash
    litellm_params:
      model: openai/deepseek-ai/DeepSeek-V4.1-Flash
      api_base: https://api.lithosai.cloud/v1
      api_key: os.environ/LITHOSAI_API_KEY
      use_chat_completions_api: true
      additional_drop_params: ["reasoning_effort", "client_metadata"]

litellm_settings:
  drop_params: true

general_settings:
  master_key: sk-codex-proxy

drop_params discards parameters a model does not accept rather than failing the request. additional_drop_params removes reasoning_effort: Codex asks for a reasoning summary, LiteLLM forwards that request in this field in a form LithosAI rejects, and Codex works without it. It also removes client_metadata, which LithosAI does not accept as a Chat Completions parameter. master_key is the token Codex sends to the proxy, not your LithosAI key.

Start the proxy

litellm --config codex.litellm.yaml --host 127.0.0.1 --port 4000

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://127.0.0.1:4000/health/liveliness

Point Codex at it

Codex configuration

~/.codex/config.toml

model = "deepseek-v4.1-flash"
model_provider = "lithosai"

[model_providers.lithosai]
name = "LithosAI"
base_url = "http://127.0.0.1:4000/v1"
env_key = "LITELLM_API_KEY"
wire_api = "responses"
requires_openai_auth = false

env_key names the variable holding the proxy’s master_key, so export that too.

export LITELLM_API_KEY="sk-codex-proxy"
codex

Troubleshooting

Symptom	Cause
`wire_api = "chat"` is no longer supported	`base_url` points at LithosAI directly. Point it at the proxy.
`404` on `/v1/responses`	The model entry uses `hosted_vllm/`. Codex needs `openai/` with `use_chat_completions_api`.
`400` with `No connected db.`	`LITELLM_API_KEY` does not match the proxy’s `master_key`. LiteLLM reports the lookup failure, not an auth error.
`400` mentioning `reasoning_effort` and `Input should be 'none', 'minimal', ...`	The proxy configuration is missing `additional_drop_params: ["reasoning_effort", "client_metadata"]`.

Issues common to every agent are listed under Common issues.

Getting Started

Models

Inference APIs

Account

Use LithosAI with Codex

Before you begin

Troubleshooting

​Before you begin

​Troubleshooting

Before you begin

Troubleshooting