Integration / Codex CLI

Sign up and sign in to websites from Codex

Connect Trusty Squire to Codex so it can finish browser setup work and use saved credentials without returning their raw values.

npx @trusty-squire/mcp@latest connect --target=codexInstallation guide →
01

Ask for the outcome

Sign in to Sentry and configure the webhook.

Codex keeps the integration goal in view while Trusty Squire works behind the login.

Set up Resend, Sentry, PostHog, and Postgres for this app.

Coordinate several provider setups from the same coding session.

Let my app call OpenAI without giving it the OpenAI key.

Create scoped app access instead of placing the provider key in the project.
02

How it works

  1. 01

    Run the Codex target

    The connect command signs this machine in and merges the squire server into ~/.codex/config.toml.

  2. 02

    Restart Codex

    Start a fresh Codex session so it discovers the newly configured MCP tools.

  3. 03

    Describe the website job

    Ask for the signup, authenticated setting, or credential outcome your code needs. Codex can then operate the website through Trusty Squire.

03

What to know

Your Codex settings stay intact

Codex reads MCP configuration from ~/.codex/config.toml. The installer parses that TOML, adds or refreshes the [mcp_servers.squire] entry, and writes the combined config back. Existing model, approval policy, sandbox, and other MCP settings are preserved by the merge.

Refresh or switch the browser identity

Rerun the connect command to refresh the MCP entry. Add --force-relogin when you need to replace a stale connected session or switch the Google or GitHub identity used by the Trusty Squire browser.

Keep the provider key out of the workspace

Codex can refer to a saved credential or ask Trusty Squire to make an authenticated call without receiving the plaintext. That keeps the raw provider key out of the transcript, generated code, and local .env file.

04