Coding agents

Coding agents speak one of three API shapes, and Odyssey serves all of them: Anthropic Messages for Claude Code, OpenAI Responses for Codex, and chat completions for most others. Any model works with any agent, because Odyssey translates between the shapes.

Create a key on the API keys page first. A dedicated key per agent, with a spend limit, keeps a runaway loop from reaching the rest of your balance.

Claude Code

Point Claude Code at Odyssey with environment variables. The base URL has no /v1: Claude Code adds it.

export ANTHROPIC_BASE_URL="https://odysseyapi.tech"export ANTHROPIC_AUTH_TOKEN="sk-ody-..."export ANTHROPIC_MODEL="anthropic/claude-sonnet-5"export ANTHROPIC_DEFAULT_HAIKU_MODEL="anthropic/claude-haiku-4.5" claude

ANTHROPIC_API_KEY works too, if you prefer it to ANTHROPIC_AUTH_TOKEN. Odyssey accepts the key in either header.

Codex CLI

Add Odyssey as a model provider using the Responses API, then export your key as ODYSSEY_API_KEY.

# ~/.codex/config.tomlmodel = "openai/gpt-5"model_provider = "odyssey" [model_providers.odyssey]name = "Odyssey"base_url = "https://odysseyapi.tech/v1"env_key = "ODYSSEY_API_KEY"wire_api = "responses"

OpenCode

Add an OpenAI-compatible provider to opencode.json and list the models you want in the picker.

{  "$schema": "https://opencode.ai/config.json",  "provider": {    "odyssey": {      "npm": "@ai-sdk/openai-compatible",      "name": "Odyssey",      "options": {        "baseURL": "https://odysseyapi.tech/v1",        "apiKey": "{env:ODYSSEY_API_KEY}"      },      "models": {        "anthropic/claude-sonnet-5": { "name": "Claude Sonnet 5" },        "openai/gpt-5": { "name": "GPT-5" }      }    }  }}

Cline and Roo Code

1. Open the extension's settings and choose OpenAI Compatible as the API provider.
2. Set the base URL to https://odysseyapi.tech/v1 and paste your Odyssey key.
3. Enter a model id from the catalogue, such as anthropic/claude-sonnet-5.

Tips for long sessions

• Agents stream, which is what they should do: very long non-streaming requests can be cut off by network timeouts before the model finishes.
• The 30 requests per minute of a new pay-as-you-go account can be tight for an agent making many small calls. Purchasing $10 of credit lifts it, and plans run from 15 to 300 requests per minute, or unlimited on Odyssey. See rate limits and plan allowances.