Docs · Start

Providers & agents

How BetterC0de connects to Claude Code, OpenAI Codex, Cursor Agent and Grok through their own CLIs — bring your own subscription, no API key required.

BetterC0de is bring-your-own-CLI. It does not resell model access, does not proxy your requests and does not need an API key of its own. Each agent runs through the official command-line tool you already have installed, with your existing login and plan.

Supported agents

Agent Vendor Auth model Notes
Claude Code Anthropic Sign in via the claude CLI (Pro / Max / API) Full tool use, permissions surfaced in BetterC0de
OpenAI Codex OpenAI ChatGPT sign-in or API key via codex Sandboxed execution modes are respected
Cursor Agent Cursor Cursor account via cursor-agent Uses your Cursor plan
Grok xAI grok CLI login

Available models and capabilities depend on your provider account. Connecting a provider does not make its paid usage free — limits, quotas and billing are exactly what the vendor gives you.

How detection works

On launch, BetterC0de looks for the supported CLIs on your PATH and checks whether they are signed in. Anything it finds appears under Settings → Providers as Ready. If a CLI is installed but not signed in, BetterC0de shows the vendor's login instructions instead of asking you for a key.

If a provider is missing:

  1. Install the vendor's CLI and sign in from a terminal once.
  2. Restart BetterC0de (it re-scans on launch).
  3. Check Settings → Providers again.

Where your data goes

Prompts, file context and tool output go directly from your machine to the provider you chose — the same path the CLI takes without BetterC0de. Their terms and data policies apply. BetterC0de's own servers only ever see an anonymous heartbeat and, if the app crashes, an error report. See Privacy & telemetry.

Using several providers at once

Every conversation picks its own provider and model. That means you can run Claude Code on a refactor while Codex writes tests and Cursor Agent reviews a PR — all on the same repository, each with its own permission prompts. Read Multi-agent workflows for how to keep them from stepping on each other.