Talaria

Talaria is a local Claude Code gateway that routes Claude to visible Codex models from your existing codex login session.

Talaria is intentionally minimal:

• No OPENAI_API_KEY mode.
• No Anthropic API passthrough.
• No OpenRouter, Ollama, or local model provider routes.

Requirements

• Python 3.10+
• Codex CLI logged into ChatGPT (codex login)
• Claude Code (claude)

Install

npm install -g @miragesystems/talaria

From source:

node bin/talaria.js

Quickstart

codex login
talaria doctor
talaria setup
talaria

In Claude Code, open /model and select one of the claude-* aliases.

Commands

• talaria – run preflight checks and launch Claude Code.
• talaria setup – run first-run diagnostics and show next actions.
• talaria doctor – detailed diagnostics.
• talaria doctor --live --model <alias> – optional live end-to-end check.
• talaria smoke – local smoke test without model usage.
• talaria smoke --live --model <alias> – optional local + upstream smoke.
• talaria serve – run gateway only (Ctrl-C to stop).

Launch flags

• --host <host> (default 127.0.0.1)
• --port <port> (default 8141)
• --dangerously-skip-permission (maps to Claude Code --dangerously-skip-permissions)
• --help (passed through to Claude when running talaria)

Security model

• Local-only by default (127.0.0.1 / localhost).
• Browser-origin and non-JSON message requests are rejected.
• Message body cap: 8 MiB.
• Codex backend is hard-pinned: https://chatgpt.com/backend-api/codex/responses.
• Token material only from local ~/.codex/auth.json.

Testing

python3 -m compileall talaria tests
python3 -m unittest discover -s tests -v
npm test

Release and docs

• CHANGELOG.md
• SECURITY.md
• CONTRIBUTING.md
• docs/RELEASE.md
• docs/TROUBLESHOOTING.md

License

MIT