Skip to main content
Codex is OpenAI’s coding agent. It stores MCP servers in config.toml and works in the CLI and the IDE extension.

Connect with sign-in

1

Add the server

Add it to ~/.codex/config.toml:
~/.codex/config.toml
Or use the command: codex mcp add spyro-ai --url https://mcp.spyro.app/mcp
2

Sign in

Sign in in your browser, choose your workspaces and click Allow.

Connect with a key

bearer_token_env_var holds the name of an environment variable, not the key. If the variable is missing when Codex starts, the server fails to connect.
~/.codex/config.toml
You can also set it with the command: codex mcp add spyro-ai --url https://mcp.spyro.app/mcp --bearer-token-env-var SPYRO_AI_MCP_KEY.

Scope it to one project

Codex can read a project’s .codex/config.toml, but only in projects you have trusted. Put the server there to share a setup with your team, and keep the key in each person’s environment.

Manage it

Inside the Codex terminal UI, /mcp shows connected servers and tools.

Try these prompts

  • “Use Spyro AI to find my critical SEO issues and open fixes for the ones in this repo.”
  • “Check Spyro AI for the keywords I am almost ranking for and update those pages’ headings.”
  • “Ask Spyro AI for my competitor keyword gaps and write a content brief.”

Tips

  • The key name in the file is snake case: mcp_servers, not mcpServers.
  • Keep the key at Read access. Raise it only for a pipeline that should make changes. Each change still needs a person to confirm it in Spyro AI.
  • If your Codex version supports tool approval settings, keep approval on for the nine manage* tools. Spyro AI also asks you to confirm each change.

If it fails

  • Server fails to start: check the environment variable is set in the same shell that starts Codex.
  • 401 error: run codex mcp login spyro-ai again, or check the key.
  • More: Troubleshooting.
OpenAI’s guide: Codex and MCP.