Integrations

Set up in two minutes.

Install the CLI or the VS Code extension, then connect your agent to the local gateway. No SDK, no code changes, no account hand-off — your keys stay yours. One gateway can serve Claude Code and an OpenAI-compatible agent at the same time.

Claude Code VS Code Cursor Cline Roo Code GitHub Copilot
0

Install TokenBeaver

CLI from npm, or the extension from the Marketplace — same gateway either way.

1

Start the gateway

CLI: run tokenbeaver-gateway in a terminal. Extension: run TokenBeaver: Start Gateway from the Command Palette (Ctrl+Shift+P / Cmd+Shift+P). Either way you get a local server and a Base URL.

> TokenBeaver: Start Gateway
TB TokenBeaver: Start Gateway ↵ run
Base URL http://127.0.0.1:4505/v1
2

Point your tool at it

Paste the Base URL into your coding tool's API settings. Keep using your own API key — TokenBeaver forwards it unchanged.

Tool
Where to paste
Claude Code (API key or subscription)
One command: tokenbeaver-gateway --install-claude-hooks — or the extension's status bar. Guide
Cursor
Settings → Models → Override OpenAI Base URL
Cline / Roo Code
Settings → API Provider → OpenAI Compatible → Base URL
GitHub Copilot
TokenBeaver Settings → Provider → GitHub Copilot (beta)
Want the full setup reference?

Provider-by-provider guides, every CLI flag and setting, dual upstream, and dedicated pages for Claude Code and GitHub Copilot — it's all in the docs.

Browse the docs→

Same tasks. Same answer. Far fewer tokens.

Without TokenBeaver
18,400
tokens / request
→
With TokenBeaver
7,600
tokens / request · −59%
Illustrative figure from internal testing. Your results vary by model and workload.