BitRouter Agent Skills

Official Agent Skill for BitRouter — a Rust LLM proxy that unifies access to OpenAI, Anthropic, Google, OpenRouter, GitHub Copilot, and OpenCode Zen/Go behind one local endpoint.

What's in this repo

One skill: /bitrouter.

skills/bitrouter/
├── SKILL.md
└── references/
    ├── cloud-setup.md
    ├── cli.md
    ├── providers.md
    ├── diagnose.md
    ├── migrate-from-litellm.md
    ├── migrate-from-openrouter.md
    ├── migrate-from-openai-compatible.md
    ├── migrate-from-anthropic-compatible.md
    ├── harness-claude-code.md
    ├── harness-codex.md
    ├── harness-hermes-agent.md
    └── harness-openclaw.md

The earlier four-skill split (bitrouter-setup, bitrouter-migrate, bitrouter-diagnose, bitrouter-providers) has been collapsed into this single skill — same coverage, less surface area for the agent to choose from, with the deep dives loaded only when needed.

Install the skill

Any agent that reads agentskills.io packages:

npx skills add BitRouterAI/agent-skills

Claude Code (manual):

cp -r skills/bitrouter ~/.claude/skills/

What /bitrouter does

Once installed, ask your agent to set up BitRouter (Local or Cloud), configure a provider, migrate off LiteLLM/OpenRouter, fix a broken proxy, or wire a coding-agent harness — the skill carries the up-to-date facts.

The skill's first move on any setup-shaped request is to ask Local or Cloud? — Cloud users may not need to install anything at all.

Use BitRouter

Cloud (managed)

1. Sign up at https://bitrouter.ai, top up credits via Stripe.
2. Mint a brk_* API key in the dashboard.
3. Point any OpenAI- or Anthropic-shaped SDK at https://api.bitrouter.ai/v1 (drop /v1 for the Anthropic SDK) with the brk_* as api_key. No local install.

Local (self-hosted)

Install the proxy:

# shell installer (macOS / Linux)
curl --proto '=https' --tlsv1.2 -LsSf \
  https://github.com/bitrouter/bitrouter/releases/latest/download/bitrouter-installer.sh | sh

# Homebrew
brew install bitrouter/tap/bitrouter

# npm
npm install -g bitrouter

Windows PowerShell:

powershell -ExecutionPolicy Bypass -c "irm https://github.com/bitrouter/bitrouter/releases/latest/download/bitrouter-installer.ps1 | iex"

Then:

export OPENAI_API_KEY=sk-...
export ANTHROPIC_API_KEY=sk-ant-...
export GEMINI_API_KEY=...

bitrouter start
# serves at http://localhost:4356

Point your SDK there:

from openai import OpenAI
client = OpenAI(base_url="http://localhost:4356/v1", api_key="unused")
client.chat.completions.create(
    model="anthropic/claude-sonnet-4-5",   # or openai/gpt-4o, google/gemini-2.5-pro, ...
    messages=[{"role": "user", "content": "hi"}],
)

Supported providers

Built-in, zero-config: openai, anthropic, google, openrouter, github-copilot, opencode-zen, opencode-go.

Plus any OpenAI-compatible endpoint (Azure OpenAI, Together, Perplexity, Groq, Ollama, LM Studio, …) added as a custom provider — see skills/bitrouter/references/providers.md.

v1.0 caveats

• Cloud is live; the CLI bridge is landing within 24h. https://api.bitrouter.ai/v1 works today with a brk_* key from the dashboard. The bitrouter login device-flow CLI command (which would mint and store the key automatically) prints "not implemented in v1.0" until the imminent ship.
• Local auth options: BYOK via env vars, per-provider OAuth (only github-copilot so far), or local virtual keys minted with bitrouter key sign.
• wallet and whoami are deferred to the separate ows workspace and the cloud bridge respectively.

Links

• Proxy: https://github.com/bitrouter/bitrouter
• Docs: https://bitrouter.ai
• Discord: https://discord.gg/G3zVrZDa5C

License

MIT — see LICENSE.