Copilot API

Reverse-engineered proxy yang bikin GitHub Copilot bisa dipake sama tools OpenAI/Anthropic

Instalasi | Cara Pakai | Arsitektur | API Reference | Konfigurasi

Heads up: Ini reverse-engineered proxy, bukan official GitHub product. Bisa aja kapan-kapan gak work. Use at your own risk ya!

Apa Itu Copilot API?

Copilot API adalah proxy server yang mentransformasi GitHub Copilot API jadi endpoint yang kompatibel sama OpenAI dan Anthropic. Lo bisa pake Copilot subscription lo sama tools kayak Cursor, Continue, Claude Code, atau aplikasi apapun yang support OpenAI/Anthropic API.

Basically, langganan GitHub Copilot lo jadi lebih "fleksibel" - bisa dipake di berbagai tools AI tanpa perlu bayar lagi.

Tech Stack

Runtime	Framework	Language	Build

Fitur Utama

Fitur	Deskripsi
Multi-API Support	OpenAI Chat Completions & Anthropic Messages API
Account Pool	Rotasi otomatis multi-akun buat ngindarin rate limit
Extended Thinking	Support Claude's adaptive thinking dengan 6 levels
Beautiful Dashboard	WebUI mobile-first buat monitoring & config
Smart Fallback	Auto-fallback ke model lain pas kena rate limit
Request Caching	LRU cache yang persist ke disk, hemat quota
Streaming Support	Full streaming support real-time
Webhook Notifications	Discord/Slack alerts buat quota low, errors, dll
Auto-Rotation	Rotasi akun otomatis based on quota/error triggers
Model Levels	Support reasoning effort levels (low sampai xhigh)
Quota Optimization	Route warmup/compact requests ke small model

Instalasi

Prerequisites

Bun >= 1.2.x
GitHub Copilot subscription (Individual/Business/Enterprise)

Quick Start

# Clone repo
git clone https://github.com/el-pablos/copilot-api.git
cd copilot-api
bun install

# Authenticate sama GitHub
bun run auth

# Start server (development)
bun run dev

# Atau production mode
bun run start

Install via npm

# Global install
npm install -g copilot-api

# atau pake bunx langsung
bunx copilot-api

Server jalan di http://localhost:4141, dashboard juga available di URL yang sama.

Cara Pakai

Konfigurasi Client

Ganti base URL di aplikasi lo:

OpenAI-compatible:

Base URL: http://localhost:4141/v1
API Key: ghu_xxxx (atau dummy)

Anthropic-compatible:

Base URL: http://localhost:4141
API Key: ghu_xxxx (atau dummy)

Dengan Claude Code

# Start server dengan opsi Claude Code
copilot-api start --claude-code

# Atau manual setup
ANTHROPIC_BASE_URL=http://localhost:4141 \
ANTHROPIC_AUTH_TOKEN=dummy \
ANTHROPIC_MODEL=gpt-4.1 \
claude

Dengan OpenAI SDK

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "http://localhost:4141/v1",
  apiKey: "dummy",
});

const response = await client.chat.completions.create({
  model: "gpt-4.1",
  messages: [{ role: "user", content: "Hello!" }],
  stream: true,
});

for await (const chunk of response) {
  process.stdout.write(chunk.choices[0]?.delta?.content || "");
}

Dengan Anthropic SDK

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  baseURL: "http://localhost:4141",
  apiKey: "dummy",
});

const response = await client.messages.create({
  model: "claude-sonnet-4-20250514",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Hello!" }],
});

console.log(response.content);

Model dengan Reasoning Level

Lo bisa pake suffix level buat kontrol reasoning effort:

# Model dengan level suffix
gpt-5.4(high)          # High reasoning effort
claude-opus-4.6(xhigh) # Extra high reasoning
gpt-5-mini(low)        # Low untuk hemat quota

# Atau pake cus- prefix buat custom request
cus-gpt-4.1            # Custom request tanpa level

Available Levels: none, minimal, low, medium, high, xhigh

Arsitektur

High-Level Architecture

flowchart TB
    subgraph Clients["Client Applications"]
        CC[Claude Code]
        CS[Cursor]
        CT[Continue]
        CA[Custom Apps]
    end

    subgraph CopilotAPI["Copilot API Proxy"]
        direction TB
        HS[Hono Server]

        subgraph Middleware["Middleware Layer"]
            CORS[CORS]
            AUTH[Auth]
            LOG[Logger]
        end

        subgraph Routes["Route Handlers"]
            OAI[OpenAI Routes]
            ANT[Anthropic Routes]
            EMB[Embeddings]
            WEB[WebUI API]
        end

        subgraph Core["Core Services"]
            CACHE[Request Cache]
            QUEUE[Request Queue]
            POOL[Account Pool]
            RATE[Rate Limiter]
        end
    end

    subgraph External["External Services"]
        GH[GitHub OAuth]
        CP[Copilot API]
    end

    Clients --> HS
    HS --> Middleware
    Middleware --> Routes
    Routes --> Core
    Core --> External

Request Pipeline Flow

flowchart LR
    A[Client Request] --> B[Hono Server]
    B --> C{Route?}

    C -->|/v1/chat/completions| D[OpenAI Handler]
    C -->|/v1/messages| E[Anthropic Handler]
    C -->|/v1/embeddings| F[Embeddings Handler]
    C -->|/v1/responses| G[Responses Handler]

    D --> H{Cache Hit?}
    E --> H
    F --> H
    G --> H

    H -->|Yes| M[Return Cached]
    H -->|No| I[Queue System]

    I --> J[Account Pool]
    J --> K[Rate Limiter]
    K --> L[Copilot API]
    L --> N[Transform Response]
    N --> O[Cache & Return]

Account Pool Strategy

flowchart TD
    A[Request Masuk] --> B{Pool Enabled?}
    B -->|No| C[Single Account]
    B -->|Yes| D{Strategy?}

    D -->|sticky| E[Same Account]
    D -->|round-robin| F[Sequential Rotation]
    D -->|quota-based| G[Select by Quota]
    D -->|hybrid| H[Sticky + Auto-Rotate]

    E --> I{Error?}
    F --> I
    G --> I
    H --> I

    I -->|Yes| J[Rotate to Next]
    I -->|No| K[Process Request]
    J --> K

    K --> L{Quota Low?}
    L -->|Yes| M[Send Webhook Alert]
    L -->|No| N[Return Response]
    M --> N

Extended Thinking Flow

flowchart TD
    A[Incoming Request] --> B{Model Check}

    B --> C{Has Thinking Support?}
    C -->|No| D[Standard Request]
    C -->|Yes| E[Check Effort Level]

    E --> F{Effort Level?}
    F -->|none| G[Budget: 0]
    F -->|minimal| H[Budget: 1024]
    F -->|low| I[Budget: 2048]
    F -->|medium| J[Budget: 4096]
    F -->|high| K[Budget: 8192]
    F -->|xhigh| L[Budget: 16384]

    G --> M[Skip Thinking]
    H --> N[Apply Thinking Budget]
    I --> N
    J --> N
    K --> N
    L --> N

    M --> O[Send to Copilot API]
    N --> O

    O --> P[Stream Response]
    P --> Q{Has Thinking Block?}
    Q -->|Yes| R[Extract & Include Thinking]
    Q -->|No| S[Standard Response]
    R --> T[Return to Client]
    S --> T

Data Flow Diagram

flowchart LR
    subgraph Storage["Data Storage"]
        CFG[("config.json")]
        TOK[("github-token.txt")]
        POOL[("pool-state.json")]
        CACHE[("request-cache.json")]
        USAGE[("usage-stats.json")]
    end

    subgraph Runtime["Runtime State"]
        STATE[State Manager]
        TOKEN[Token Manager]
        ACCT[Account Pool]
        RCACHE[Request Cache]
    end

    CFG --> STATE
    TOK --> TOKEN
    POOL --> ACCT
    CACHE --> RCACHE

    STATE --> |"Config changes"| CFG
    TOKEN --> |"Refresh token"| TOK
    ACCT --> |"Pool updates"| POOL
    RCACHE --> |"Cache persist"| CACHE
    USAGE --> |"Track usage"| STATE

Struktur Project

copilot-api/
├── src/
│   ├── main.ts                # CLI entry point (citty)
│   ├── server.ts              # Hono app + middleware setup
│   ├── start.ts               # Server bootstrap & init
│   ├── auth.ts                # GitHub OAuth flow
│   │
│   ├── lib/                   # Core utilities
│   │   ├── account-pool.ts           # Multi-account management
│   │   ├── account-pool-quota.ts     # Quota tracking
│   │   ├── account-pool-notify.ts    # Webhook notifications
│   │   ├── config.ts                 # File-based config
│   │   ├── request-cache.ts          # LRU caching
│   │   ├── request-queue.ts          # Concurrent handling
│   │   ├── reasoning.ts              # Thinking utilities
│   │   ├── model-level.ts            # Model level parsing
│   │   ├── token.ts                  # Token management
│   │   └── state.ts                  # Runtime state
│   │
│   ├── routes/                # API endpoints
│   │   ├── chat-completions/         # OpenAI /v1/chat/completions
│   │   ├── messages/                 # Anthropic /v1/messages
│   │   ├── embeddings/               # OpenAI /v1/embeddings
│   │   ├── models/                   # GET /models
│   │   ├── responses/                # OpenAI Responses API
│   │   └── ...
│   │
│   ├── services/              # External services
│   │   ├── copilot/                  # GitHub Copilot API client
│   │   └── github/                   # GitHub OAuth & API
│   │
│   └── webui/                 # Dashboard API routes
│
├── public/                    # WebUI frontend (Alpine.js + Tailwind)
├── tests/                     # Test files
└── dist/                      # Build output

API Reference

OpenAI Endpoints

Endpoint	Method	Deskripsi
`/v1/chat/completions`	POST	Chat completion (streaming/non-streaming)
`/v1/embeddings`	POST	Text embeddings
`/v1/models`	GET	List available models
`/v1/models/:id`	GET	Get specific model info
`/v1/responses`	POST	OpenAI Responses API

Anthropic Endpoints

Endpoint	Method	Deskripsi
`/v1/messages`	POST	Anthropic Messages API

Utility Endpoints

Endpoint	Method	Deskripsi
`/health`	GET	Health check
`/usage`	GET	Usage statistics
`/token`	GET	Current Copilot token info
`/account-limits`	GET	Account quota/limits

WebUI API

Endpoint	Method	Deskripsi
`/`	GET	Dashboard
`/api/config`	GET/POST	Get/update configuration
`/api/accounts`	GET/POST	List/add pool accounts
`/api/accounts/:id`	DELETE	Remove account from pool
`/api/cache/stats`	GET	Cache statistics
`/api/cache/clear`	POST	Clear cache
`/api/logs/stream`	GET	Real-time log stream (SSE)
`/api/notifications/stream`	GET	Notification stream (SSE)

Konfigurasi

Config file ada di ~/.config/copilot-api/config.json

Environment Variables

Variable	Default	Deskripsi
`PORT`	`4141`	Server port
`DEBUG`	`false`	Debug logging
`GH_TOKEN`	-	GitHub token
`WEBUI_PASSWORD`	-	Password buat WebUI
`HTTP_PROXY`	-	HTTP proxy URL
`HTTPS_PROXY`	-	HTTPS proxy URL
`FALLBACK`	`false`	Enable model fallback
`CHAT_COMPLETION_TIMEOUT_MS`	`300000`	Request timeout (5 menit)

CLI Options

copilot-api start [options]

Options:
  -p, --port <port>           Port (default: 4141)
  -v, --verbose               Verbose logging
  -d, --debug                 Debug mode
  -g, --github-token <token>  GitHub token langsung
  -c, --claude-code           Generate Claude Code command
  -f, --fallback              Enable model fallback
  --proxy-env                 Pake HTTP_PROXY/HTTPS_PROXY dari env
  --webui-password <pass>     Set WebUI password

Account Pool Strategies

Strategy	Deskripsi	Kapan Pake
`sticky`	Pake akun yang sama sampe error	Default, simple usage
`round-robin`	Rotasi berurutan tiap request	Load balancing rata
`quota-based`	Pilih berdasarkan sisa quota	Maximize quota usage
`hybrid`	Sticky + auto-rotate pas error	Recommended!

Contoh Config Lengkap

{
  "port": 4141,
  "debug": false,
  "apiKeys": [],

  "poolEnabled": true,
  "poolStrategy": "hybrid",
  "poolAccounts": [],

  "cacheEnabled": true,
  "cacheMaxSize": 1000,
  "cacheTtlSeconds": 3600,

  "queueEnabled": true,
  "queueMaxConcurrent": 3,
  "queueMaxSize": 100,
  "queueTimeout": 60000,

  "fallbackEnabled": false,
  "modelMapping": {},

  "autoRotationEnabled": true,
  "autoRotationTriggers": {
    "quotaThreshold": 10,
    "errorCount": 3,
    "requestCount": 0
  },
  "autoRotationCooldownMinutes": 30,

  "webhookEnabled": false,
  "webhookProvider": "discord",
  "webhookUrl": "",
  "webhookEvents": {
    "quotaLow": { "enabled": true, "threshold": 10 },
    "accountError": true,
    "rateLimitHit": true,
    "accountRotation": true
  },

  "modelReasoningEfforts": {
    "gpt-5-mini": "low",
    "gpt-5.3-codex": "xhigh",
    "gpt-5.4-mini": "xhigh",
    "gpt-5.4": "xhigh",
    "claude-opus-4.5": "xhigh",
    "claude-sonnet-4.5": "xhigh"
  },

  "smallModel": "gpt-5-mini",
  "compactUseSmallModel": true,
  "warmupUseSmallModel": true,

  "defaultMaxOutputTokens": 32768,
  "maxContextTokensOverride": 0,
  "disableTruncation": false,
  "claudeTokenMultiplier": 1.15,

  "requestTimeoutMs": 300000,
  "trackUsage": true,
  "trackCost": true
}

Release, Tagging, dan CI/CD

Project ini pake alur rilis otomatis supaya setiap perubahan di main bisa langsung punya versi/tag yang jelas.

CI utama: .github/workflows/ci.yml
- jalan saat push ke main/develop dan saat PR ke main
- gate wajib: lint + typecheck + test + build
Auto tag: dari job auto-tag di workflow CI
- bikin tag versi terbaru pas commit masuk main
- kalau tag versi sudah ada, versi patch akan dibump otomatis sebelum tag baru dibuat
Release latest: .github/workflows/release.yml
- jalan saat ada tag v*.*.*
- generate changelog dan publish GitHub Release dengan status latest
Manual bump: .github/workflows/version-bump.yml
- bisa dipicu manual (workflow_dispatch) untuk bump patch/minor/major

Flow singkat:

flowchart LR
    A[Push ke main] --> B[CI: lint typecheck test build]
    B --> C[Auto Tag vX.Y.Z]
    C --> D[Trigger Release Workflow]
    D --> E[Generate changelog]
    E --> F[Publish GitHub Release latest]

Development

bun run dev        # Development server (hot reload)
bun run build      # Build project
bun test           # Run tests
bun run lint       # Lint code
bun run typecheck  # Type check

Code Style

Imports: Pake ~/* alias buat src/*
Types: Strict TypeScript, no any
Naming: camelCase buat variables, PascalCase buat types
Modules: ESNext only, no CommonJS

Data Storage

Path	Deskripsi
`~/.config/copilot-api/config.json`	Configuration
`~/.config/copilot-api/request-cache.json`	Request cache
`~/.local/share/copilot-api/github-token.txt`	GitHub token
`~/.local/share/copilot-api/pool-state.json`	Account pool state
`~/.local/share/copilot-api/usage-stats.json`	Usage statistics

Troubleshooting

Token expired

copilot-api auth

Rate limit

Enable multi-account pool di dashboard
Pake hybrid strategy
Enable request queue
Tambah lebih banyak akun ke pool

Model not found

Check available models: GET /models
Enable fallbackEnabled: true
Cek model mapping di config

Cache issues

curl -X POST http://localhost:4141/api/cache/clear

Kontributor

Makasih buat semua yang udah contribute ke project ini!

Cara Contribute

Fork repo ini
Buat branch baru: git checkout -b feature/nama-fitur
Commit changes: git commit -m "add: deskripsi singkat"
Push ke branch: git push origin feature/nama-fitur
Buat Pull Request

License

MIT License - lihat LICENSE

Made with yang penting works

Name		Name	Last commit message	Last commit date
Latest commit History 337 Commits
.github/workflows		.github/workflows
.vscode		.vscode
backup-files		backup-files
docs		docs
e2e		e2e
fixed-files/01-round-robin-bug		fixed-files/01-round-robin-bug
pages		pages
public		public
screenshots		screenshots
src		src
tests		tests
.gitignore		.gitignore
AGENTS.md		AGENTS.md
ANALISIS_PERBANDINGAN_CINA_COPILOT_VS_COPILOT_API.md		ANALISIS_PERBANDINGAN_CINA_COPILOT_VS_COPILOT_API.md
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
MEGA_PROMPT_IMPLEMENTASI.md		MEGA_PROMPT_IMPLEMENTASI.md
README.md		README.md
bun.lock		bun.lock
bunfig.toml		bunfig.toml
cliff.toml		cliff.toml
eslint.config.js		eslint.config.js
opencode.json		opencode.json
package.json		package.json
playwright.config.ts		playwright.config.ts
start.bat		start.bat
tailwind.config.js		tailwind.config.js
tsconfig.json		tsconfig.json
tsdown.config.ts		tsdown.config.ts

Folders and files

Latest commit

History

Repository files navigation

Copilot API

Apa Itu Copilot API?

Tech Stack

Fitur Utama

Instalasi

Prerequisites

Quick Start

Install via npm

Cara Pakai

Konfigurasi Client

Dengan Claude Code

Dengan OpenAI SDK

Dengan Anthropic SDK

Model dengan Reasoning Level

Arsitektur

High-Level Architecture

Request Pipeline Flow

Account Pool Strategy

Extended Thinking Flow

Data Flow Diagram

Struktur Project

API Reference

OpenAI Endpoints

Anthropic Endpoints

Utility Endpoints

WebUI API

Konfigurasi

Environment Variables

CLI Options

Account Pool Strategies

Contoh Config Lengkap

Release, Tagging, dan CI/CD

Development

Code Style

Data Storage

Troubleshooting

Token expired

Rate limit

Model not found

Cache issues

Kontributor

Cara Contribute

License

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 2

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages