Which config file does Codex use?

Use ~/.codex/config.toml for the recommended WaveSpeed AI setup. The configuration points Codex to endpoint /v1/responses.

Which protocol should Codex use?

Codex should use OpenAI Responses API. This page uses endpoint /v1/responses to match that protocol.

How do I verify Codex is connected?

Run Codex locally and open WaveSpeed Logs. Successful requests should appear with the selected model and usage details.

Agent setup

Connect Codex to WaveSpeed LLM API

Copy the recommended ~/.codex/config.toml setup for Codex, route requests to /v1/responses, and verify local agent calls in WaveSpeed Logs.

API key Test model

Config file

~/.codex/config.toml

Protocol

OpenAI Responses API

Endpoint

/v1/responses

Quick start

Create a WaveSpeed access key

Export the required key locally

Paste config into~/.codex/config.toml

Run codex --profile wavespeed and check Logs

Model

OpenAI-compatible chat completions

Usage appears in Dashboard and Logs

Switch models without rewriting agent code

Edit ~/.codex/config.toml

# File: ~/.codex/config.toml

model = "anthropic/claude-opus-4.8"
model_provider = "wavespeed"

[model_providers.wavespeed]
name = "WaveSpeed AI"
base_url = "https://llm.wavespeed.ai/v1"
env_key = "WAVESPEED_API_KEY"
wire_api = "responses"

Quick start

Create a WaveSpeed access key

Export the required key locally

Paste config into~/.codex/config.toml

Run codex --profile wavespeed and check Logs

Environment

export WAVESPEED_API_KEY="sk-..."

After setup

View usage dashboard View request logs Browse models

Agent integration guide

Use Codex with WaveSpeed LLM API

This guide shows how to connect Codex to WaveSpeed AI with ~/.codex/config.toml, OpenAI Responses API, /v1/responses, base URL https://llm.wavespeed.ai/v1, and default model anthropic/claude-opus-4.8. Copy the configuration above, export your API key, run the agent, then verify requests in Logs.

Create a WaveSpeed access key

Export the required key locally

Paste config into

~/.codex/config.toml

Run codex --profile wavespeed and check Logs

What this Codex setup configures

Protocol

OpenAI Responses API

Endpoint

/v1/responses

Model

anthropic/claude-opus-4.8

The copy-ready configuration above is the only Codex file you need to edit for the recommended setup: ~/.codex/config.toml. Keep the API key in your shell environment so it is not committed to a repository.

Troubleshooting

The API key is not being picked up

Export WAVESPEED_API_KEY in the same shell where you run Codex. If you use a shell profile, reload the terminal before starting the agent.

Requests fail with a protocol or endpoint error

Check that Codex is using OpenAI Responses API and endpoint /v1/responses. Protocol mismatches are the most common cause of 404 or payload errors.

The selected model does not work

Confirm the model id exists in WaveSpeed Models, then paste the exact id into ~/.codex/config.toml. You can start with the default model shown on this page.

No requests appear in Logs

Make sure Codex is using the WaveSpeed base URL and your current API key. Then send a new prompt from the local agent and refresh Logs.

Can this Codex guide show up in search?

Yes. This page is built as an indexable setup guide for users searching for Codex custom provider, OpenAI-compatible endpoint, base URL, and local agent configuration workflows.

View all agent setup guides Browse models