OpenBulu
Taking longer than expected? Reload this page.

Codex CLI

Prefer the Responses configuration, usually with the Base URL ending at /v1. Codex-related requests may also use /v1/responses/compact, so validate with a small request first.

Protocol: OpenAI Responses. Base URL: https://openbulu.com/v1. Endpoint: POST /v1/responses.

Configuration example

model = "<MODEL_ID>"
model_provider = "openbulu"
approval_policy = "on-request"
sandbox_mode = "workspace-write"

[model_providers.openbulu]
name = "OpenBulu"
base_url = "https://openbulu.com/v1"
env_key = "OPENBULU_API_KEY"
wire_api = "responses"
requires_openai_auth = false
supports_websockets = false

# Windows only, after the MXC sandbox probe succeeds
[windows]
sandbox = "mxc"

Examples use placeholders only. Keep your real API key in an environment variable or the client’s secure credential store.

Setup steps

  1. Install Codex CLI and save your OpenBulu API key in the OPENBULU_API_KEY environment variable.
  2. Add the OpenBulu provider to ~/.codex/config.toml, select your model ID, and set wire_api to responses.
  3. For the first test, start Codex in a disposable folder and run /status. Confirm the OpenBulu provider, selected model, and Workspace (Ask for approval), then send the connection test prompt. For file-edit tasks, review changes and run the test independently.

Verify setup

Create a key, choose a model available to your account, run a small request in the client, then check its model, result, and charge in Usage Logs. Do not share the full key.

The connection prompt gets the exact reply, and its usage log shows /v1/responses plus the selected model. If you also test file editing, keep that work in a disposable folder and run its test independently.

Troubleshooting

All client guides · Model pricing