How to Connect ChatGPT Codex to CoCoCo

Before you start

• Codex CLI installed (the IDE extension shares the same config)
• Your Endpoint URL and an active API Token from the MCP Connection page (see How to Find Your MCP Connection Details)

Step 1 — Put your token in an environment variable

Codex sends the token from an environment variable rather than storing it in the config file:

export COCOCO_TOKEN="your-api-token"

Set it in the shell (or shell profile) where you launch Codex.

Step 2 — Add the CoCoCo server to config.toml

Codex stores MCP servers in ~/.codex/config.toml (global) or .codex/config.toml (trusted project). Add:

[mcp_servers.cococo]
url = "https://<your-domain>/mcp"
bearer_token_env_var = "COCOCO_TOKEN"

bearer_token_env_var tells Codex to send Authorization: Bearer <token> using the variable from Step 1.

Or add it from the CLI:

codex mcp add cococo --url https://<your-domain>/mcp --bearer-token-env-var COCOCO_TOKEN

Step 3 — Verify

Run codex, then type /mcp to see the configured servers. CoCoCo should appear as connected with its tools.

Troubleshooting

• Server missing or empty: make sure COCOCO_TOKEN is exported in the same shell that launches Codex.
• Remote server not picked up: some Codex versions need the streamable-HTTP client enabled — add experimental_use_rmcp_client = true to config.toml.
• Authentication errors: confirm the token is active and not revoked.
• The CLI and the IDE extension share config.toml, so you only configure this once.

Related articles

• How to Find Your MCP Connection Details