Copilot API
Universal AI Gateway
One Gateway. Any Client. Multiple AI Providers.
Chat Completions · OpenAI Responses · Anthropic Messages
English | 简体中文
Copilot API is a local AI gateway that connects Claude Code, OpenCode, Codex, and other clients to GitHub Copilot, the built-in Codex and xAI providers, and third-party model providers through a unified API.
Highlights
- Unified API Gateway: Serve OpenAI-compatible Chat Completions (
/v1/chat/completions), the OpenAI Responses API (/v1/responses), and Anthropic-compatible Messages (/v1/messages) from one local endpoint.
- Multi-Provider: Route GitHub Copilot, the built-in
codex and xai providers, and third-party providers (Kimi, DeepSeek, DashScope, OpenRouter, OpenCode Go, Google Gemini, or a custom provider from models.dev) behind the same gateway. GitHub Copilot is optional — with at least one enabled provider, the server starts in provider-only mode without a GitHub token.
- Coding Agent Ready: First-class setups for Claude Code, OpenCode, and Codex, including settings.json gateway model discovery and a merged model catalog for Codex.
- Streaming & WebSocket: SSE streaming on all three client-facing protocols. Upstream Copilot Responses traffic selects WebSocket or HTTP from each model’s advertised endpoints; streamed Responses traffic for the built-in
codex provider uses WebSocket by default and uses HTTP when useResponsesApiWebSocket is disabled.
- Desktop App: Electron GUI with GitHub Copilot sign-in, Codex and xAI OAuth, provider configuration, token usage, logs, and one-click start/stop.
Quick Start
Use Node.js 22.x, 24.x, or 26.x with the security updates below (npx) or Bun >= 1.4.4. A Copilot subscription is needed only for the GitHub Copilot provider; other configured providers can run independently.
CVE-2026-111497: The Brotli decoder in affected Node.js and Bun runtimes can corrupt memory and allow remote code execution. Upgrade Node.js 22.x to 22.23.4 or later, 24.x to 24.22.1 or later, or 26.x to 26.12.1 or later; upgrade Bun to 1.4.4 or later. Restart the gateway after upgrading. For Docker deployments, rebuild and redeploy the image with a patched runtime. See the Node.js security release and Bun security advisory.
npx @jeffreycao/copilot-api@latest start
The server listens on http://localhost:4141 by default. Optionally authenticate with GitHub Copilot or configure a third-party provider first:
npx @jeffreycao/copilot-api@latest auth login
Alternatively, run with Bun >= 1.4.4, which can offer better performance:
bunx --bun @jeffreycao/copilot-api start
Use --bun to select the Bun runtime and omit @latest for routine starts to reuse the cache. See Bun setup for global installation and upgrades.
[!NOTE]
To trust system-installed CA certificates (for example, behind a corporate HTTPS proxy), npx needs no extra configuration: Node.js >= 22.19.0 merges the system CA store at startup through tls.setDefaultCACertificates(). Only Bun requires NODE_USE_SYSTEM_CA=1 before bunx --bun: export NODE_USE_SYSTEM_CA=1 in bash/zsh, or $env:NODE_USE_SYSTEM_CA="1" in PowerShell (Bun >= 1.4.4).
Desktop enables the variable for server subprocesses by default, preserving explicit environment settings.
Verify the gateway is up:
curl http://localhost:4141/v1/models
Token usage storage requires Bun >= 1.4.4 or Node.js with built-in node:sqlite (22.13.0+). Use a patched Node.js release from the warning above. See Using with npx for details.
From here, jump to the guide for your client: Claude Code, OpenCode, Codex, DeepSeek Harness, or run it with Docker.
Codex Multi-Agent compatibility: codexMultiAgentV1 optionally advertises V1; V2 remains the default. Apply the gateway setting and start a new Codex session. Catalog refreshes do not switch an already-selected session version, and resumed sessions retain their recorded version. Explicit CLI overrides may take precedence.
Compatibility
Every client talks to the same local endpoint. The gateway routes each request to GitHub Copilot, the built-in codex or xai provider, or a configured third-party provider, translating between protocols when the provider speaks a different one.
Client / Protocol Matrix
| Client |
Chat Completions |
Responses |
Anthropic Messages |
Recommended |
| Claude Code |
— |
— |
✅ Native / Adapter |
Anthropic Messages |
| OpenCode |
✅ Native |
✅ Native / Adapter |
✅ Native / Adapter |
Anthropic Messages |
| Codex |
— |
✅ Native / Adapter |
— |
Responses |
| DeepSeek Harness |
— |
✅ Native / Adapter |
✅ Native / Adapter |
Responses |
| OpenAI-compatible clients |
✅ Native |
✅ Native / Adapter |
— |
Chat Completions |
| Anthropic-compatible clients |
— |
— |
✅ Native / Adapter |
Anthropic Messages |
Providers and protocols. Protocol support is model-specific. Chat Completions requires a native endpoint, while Responses and Messages can use supported adapters. The built-in codex provider uses Responses natively; third-party providers can use anthropic, openai-compatible, or openai-responses, with per-model overrides.
Desktop App
Prefer a GUI? The Electron desktop app in desktop/ covers GitHub Copilot sign-in, OpenAI Codex and xAI OAuth, and API-key configuration for Kimi, DeepSeek, DashScope, OpenRouter, or a custom provider from models.dev — with one-click start/stop of the local server, and the local endpoint, auth header, available models, usage, and logs in one window.
Windows x64 (.exe), macOS Apple Silicon (.dmg), and Linux x64 (.AppImage) packages are published in GitHub Releases. See Electron Desktop App for full setup and advanced configuration.
After the initial full-package installation, subsequent updates normally use differential downloads to reduce download size. On macOS, the first update may still download a complete DMG to establish the cache. If the cache is unavailable or differential downloading fails, the app falls back to a full download.
Documentation
| Guide |
Contents |
| Installation and Startup |
Prerequisites, project overview, npx and source runs, provider-only mode without Copilot, and gateway API key setup |
| Claude Code |
settings.json, gateway model discovery, model mappings, auto-compact window, and WebSearch behavior |
| OpenCode |
v2 Responses plugin and automatic model discovery, native compaction, model capabilities and pricing, context limits, and v1 Anthropic SDK setup |
| Codex |
A full config.toml provider block, GITHUB_COPILOT_API_KEY environment variable setup, auto-review model mapping, generating model_catalog.json, and the merged model picker catalog with protocol adapters |
| DeepSeek Harness |
Install dsh-copilot-api, save the gateway API key, and choose Responses or Messages |
| Docker |
Docker Compose quick start, the /data persistent mount and its ownership repair, supported environment variables, and host interface binding |
| Desktop App |
Copilot sign-in, Codex OAuth account switching, API-key providers, one-click start / stop, shared model mappings, advanced settings, and per-platform installers |
| Plugins and Tool Search |
The Responses tool_search MCP bridge (not needed on opencode v2, which already defers tools through Code Mode), Claude Code agent-inject and tool-search marketplace plugins, and the opencode subagent marker plugin |
| Usage Monitoring |
The usage viewer URL and query parameters, period selectors, Copilot quota progress, token and cost metric cards, trend charts, and paginated request events |
| CLI Reference |
Command structure, global options, and the full option sets for the start, auth, and debug subcommands with example usage |
| Configuration Reference |
Every config.json field: gateway and admin API keys, provider definitions, model mappings, WebSocket and HTTP transport, timeouts, and context management |
| API and Authentication |
Allowed auth headers and CORS rules, OpenAI, Codex backend, and Anthropic endpoints, usage monitoring routes, and admin configuration endpoints |
| Troubleshooting |
Fixes for Claude models missing from the Copilot model list |
Before using GitHub Copilot, read the GitHub Copilot Security Notice.
评论