⚡ OpenCode2API v3.2.2

High-Performance OpenAI & Anthropic Compatible Gateway for OpenCode Models

Zero-config intelligent egress: discovers, verifies, latency-ranks, rotates, and prunes public HTTP/HTTPS routes with Bun TLS fingerprint synchronization, resilient proxy failover, and real-time SSE streaming.

Warning

⚠️ Educational & Research Purpose Disclaimer

This software project is designed and released strictly for educational, protocol research, and reverse-engineering study purposes. It demonstrates modern API translation architectures, loopback TLS fingerprint matching, dynamic proxy pool lifecycle management, and cross-schema streaming protocols between OpenAI SSE and Anthropic Messages.

The developers and contributors do not encourage, condone, or support any abuse of third-party platforms or violation of upstream Terms of Service. Users are exclusively responsible for their own usage and must ensure compliance with all applicable platform policies, terms of service, and local laws.

📖 Table of Contents

What OpenCode2API Does
Architecture & Upstream Quota Reality
- Residential ISP vs. Cloud Datacenter IPs
- 3 Ways to 100% Reliability
Key Features in v3.1.0
Discovered Models Roster
Quickstart: Local Installation (Recommended)
Quickstart: Docker & Docker Compose
Deploying to Dokploy / Coolify
Zero-Cost Remote Access via Cloudflare Tunnel
Turnkey PowerShell Testing Scripts
API Usage Examples
Configuration Reference
License

What OpenCode2API Does

OpenCode2API acts as a drop-in gateway translating standard OpenAI (/v1/chat/completions) and Anthropic (/v1/messages) requests to OpenCode Zen inference endpoints. It eliminates the friction of proprietary CLI formats, offering:

Universal Client Support: Works natively in Cursor, VS Code (Continue/Cline), LibreChat, OpenAI SDK, Anthropic SDK, LangChain, and LiteLLM.
Deep Reasoning Traces: Streams interleaved <think> blocks as reasoning_content (OpenAI) and thinking_delta (Anthropic).
Core Tool Calling: Merges OpenCode CLI tool schemas upstream while allowing custom client function calling.
Vision Integration: Accepts multi-modal images via OpenAI base64 image_url and Anthropic image blocks.
Built-in Control Plane: React-based dashboard with live request telemetry, interactive playground, scoped O2P- API key manager, and proxy diagnostics.

Architecture & Upstream Quota Reality

Residential ISP vs. Cloud Datacenter IPs

When using the anonymous free tier (Authorization: Bearer public), OpenCode enforces per-IP rate limiting and daily usage quotas guarded by Cloudflare bot management:

┌────────────────────────────────────────────────────────────────────────┐
│                        UPSTREAM EGRESS ARCHITECTURE                    │
├───────────────────────────────┬────────────────────────────────────────┤
│ Local PC (Residential ISP)    │ Cloud VPS / Public Datacenter Proxies   │
├───────────────────────────────┼────────────────────────────────────────┤
│ ✅ Clean Residential IP Range │ ❌ Shared Datacenter ASN (AWS/Hetzner) │
│ ✅ Never flagged by Bot Guard │ ❌ Cloudflare IP reputation penalty    │
│ ✅ Full Free Daily Quota      │ ❌ Shared Free Quota exhausted early   │
│ ✅ 100% Success Rate (2–4s)   │ ⚠️ Returns FreeUsageLimitError (429)   │
└───────────────────────────────┴────────────────────────────────────────┘

When public proxies from scraped lists are used on cloud servers (Dokploy, VPS, AWS), Cloudflare detects datacenter ASN ranges where multiple bots have already burned the anonymous free tier quota for that day, returning:

{
  "type": "error",
  "error": {
    "type": "FreeUsageLimitError",
    "message": "Free usage limit reached. Free models work in OpenCode (https://opencode.ai); subscribe to OpenCode Go to keep going: https://opencode.ai/go"
  }
}

(With an upstream Retry-After: 41514 header resetting at midnight UTC).

3 Ways to 100% Reliability

Depending on your environment, you can achieve 100% zero-drop pass rates:

Dedicated Rotating Proxy (Single Endpoint): Add a single rotating proxy endpoint (Webshare, Smartproxy, Oxylabs, BrightData, IPRoyal) under Settings -> Rotating Proxy Gateway. Each request exits through a different residential/mobile IP automatically, eliminating IP rate limits and avoiding 60s quota lockouts without maintaining proxy lists.
Local Daemon (Zero Cost, 100% Free): Running OpenCode2API on your local machine uses your personal broadband/residential IP. It bypasses Cloudflare datacenter blocks with 100% success rate across all models.
Upstream Zen Key (Recommended for Cloud VPS / Dokploy): Add your genuine OpenCode Zen Key / API key in Dashboard Settings or via OPENCODE_ZEN_KEY environment variable. Authenticated requests use Bearer <key>, completely bypassing anonymous IP quotas on any server.
Scraped / Manual US Proxy Pool: If hosting on a VPS without a Zen key or rotating proxy, enable automatic discovery or paste manual proxy URLs under Proxy Egress -> Manual Proxies.

Key Features in v3.1.0

🔄 Dedicated Rotating Proxy Support: Add a single rotating proxy endpoint in the dashboard instead of managing multi-line lists. Auto-injects dynamic session identifiers and closes TCP keep-alive sockets per request, forcing fresh exit IPs and eliminating 429 quota lockouts.
⚡ Ultra-Strict 3s High-Throughput Scraper & 5x Verifier: Hard 3000ms screening timeout, 5 consecutive verification checks, and parallelized multi-threaded screening (200 workers) and verification (60 workers) across high-yield public sources.
🚀 Instant Rate-Limit & Quota Switching: Zero-latency proxy hopping on upstream 429 or quota limits without artificial delays or hangs. Individual proxy timeouts bounded to prevent client hanging.
🧵 Loopback Bun TLS Bridge: OpenCode's upstream gate rejects Node.js TLS fingerprints. v3.1.0 launches an embedded Bun worker under BUN_BE_BUN=1 using OpenCode CLI v2.1.4, matching upstream client fingerprints byte-for-byte.
⚡ Latency-Ranked Routing: Active proxies are sorted dynamically by measured latency ascending; completions round-robin exclusively across the top 12 fastest routes.
🛡️ Pool-First Sequential Failover: When a proxy hits an upstream rate limit, it cools down and immediately fails over to the next fastest proxy in the pool (up to 20 failovers) before attempting fallback.
🌐 13 US-Targeted Proxy Sources: Continuously screens candidates targeting United States endpoints to eliminate RegionError: This model is not available in your country (403).
🔄 Bun Proxy Protocol Safety: Enforces pure HTTP/HTTPS CONNECT tunneling across all scraped candidates, eliminating UnsupportedProxyProtocol crashes.
📊 No Synthetic Telemetry: Real 60-second sliding-window RPM, TPM, latency, and token allocation calculations with detailed request audit logs.

Discovered Models Roster

Discovered and dynamically mapped from OpenCode Zen:

Model ID	Provider	Context	Modality	Reasoning
`big-pickle`	OpenCode Zen	128k	Text	✅ Extended
`beam`	OpenCode Zen	128k	Text	✅ Native
`glm-5.3-flash`	GLM / Zhipu	128k	Text / Code	✅ Native
`deepseek-v4.1-flash`	DeepSeek	128k	Text / Code	✅ Native
`mimo-v2.6-flash`	Mimo	128k	Multimodal Vision	❌ Standard
`nemotron-3.5-super-vl`	NVIDIA	128k	Multimodal Vision	❌ Standard
`nemotron-3.5-lightning`	NVIDIA	128k	Text / Code	❌ Fast
`flint-40b`	Flint	128k	Text / Code	❌ Standard

Quickstart: Local Installation (Recommended)

Running locally gives you 100% free, unthrottled access through your residential connection.

1. Prerequisites

Node.js: v20 or higher installed (nodejs.org)
OpenCode CLI: Installed globally for the Bun TLS engine:
```
npm install -g @opencode/cli@2.1.4
```

2. Clone & Install

git clone https://github.com/Nystic-Shadow/Opencode2api.git
cd Opencode2api
npm install
npm run build

3. Start Gateway

# Windows PowerShell
$env:DASHBOARD_PASS="762920"; npm start

# Linux / macOS
DASHBOARD_PASS="762920" npm start

4. Access Gateway & Playground

Dashboard & Playground: http://127.0.0.1:8080/
API Base URL: http://127.0.0.1:8080/v1
Master API Key: View in data/keys.json or dashboard.

Quickstart: Docker & Docker Compose

Deploy a self-contained container with Node.js 20 and OpenCode CLI bundled:

# 1. Clone repository
git clone https://github.com/Nystic-Shadow/Opencode2api.git
cd Opencode2api

# 2. Configure environment
cp .env.example .env

Edit .env:

PORT=8080
DASHBOARD_USER=admin
DASHBOARD_PASS=MySecretPassword123!
EGRESS_MODE=auto
AUTO_PROXY_TARGET=100
ALLOW_DIRECT_FALLBACK=true

Start the stack:

docker compose up -d --build

Deploying to Dokploy / Coolify

OpenCode2API deploys seamlessly on Dokploy or Coolify:

Create a new Application in Dokploy from GitHub repository: https://github.com/Nystic-Shadow/Opencode2api.git (Branch: main).
Build Type: Dockerfile (Port: 8080).

Set Environment Variables:

DASHBOARD_USER=shadow
DASHBOARD_PASS=762920
EGRESS_MODE=auto
AUTO_PROXY_TARGET=100
ALLOW_DIRECT_FALLBACK=true
OPENCODE_ZEN_KEY=your_optional_zen_key_here

Click Deploy. Dokploy will build the image, initialize the Bun TLS bridge, and start serving.

Zero-Cost Remote Access via Cloudflare Tunnel

Want to access your local residential gateway from your phone, laptop, or cloud apps without port-forwarding or public IP exposure?

Run Cloudflare Tunnel directly on your local machine:

npx cloudflared tunnel --url http://127.0.0.1:8080

Cloudflare generates a secure, free HTTPS URL (e.g. https://random-subdomain.trycloudflare.com). You now have an internet-accessible gateway powered by your clean residential IP!

Turnkey PowerShell Testing Scripts

Ready-to-use PowerShell verification scripts are included in scripts/:

1. Test Chat Completions (Any Model)

powershell -ExecutionPolicy Bypass -File scripts\test-chat.ps1 -Model "big-pickle"

(Supports -Model "beam", -Model "glm-5.3-flash", -Model "deepseek-v4.1-flash").

2. Test Real-Time SSE Token Streaming

Streams characters live to console and calculates Time-To-First-Token (TTFT) and token velocity:

powershell -ExecutionPolicy Bypass -File scripts\test-stream.ps1 -Model "glm-5.3-flash"

API Usage Examples

OpenAI Chat Completions

curl http://127.0.0.1:8080/v1/chat/completions \
  -H "Authorization: Bearer O2P-54ce9d315cbf95864a41accc5e2fc69e" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "big-pickle",
    "messages": [
      {"role": "user", "content": "Explain quantum superposition in 20 words."}
    ],
    "stream": false
  }'

Anthropic Messages

curl http://127.0.0.1:8080/v1/messages \
  -H "x-api-key: O2P-54ce9d315cbf95864a41accc5e2fc69e" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.3-flash",
    "max_tokens": 150,
    "messages": [
      {"role": "user", "content": "Write a haiku about high-speed AI gateways."}
    ]
  }'

Streaming in Python

from openai import OpenAI

client = OpenAI(
    base_url="http://127.0.0.1:8080/v1",
    api_key="O2P-54ce9d315cbf95864a41accc5e2fc69e"
)

response = client.chat.completions.create(
    model="glm-5.3-flash",
    messages=[{"role": "user", "content": "Count from 1 to 10."}],
    stream=True
)

for chunk in response:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
print()

Configuration Reference

Key environment variables in .env:

Variable	Default	Description
`PORT`	`8080`	Port for the gateway server
`DASHBOARD_USER`	`admin`	Admin dashboard username
`DASHBOARD_PASS`	Auto-generated	Admin dashboard password (min 12 chars)
`OPENCODE_ZEN_KEY`	`""`	Optional OpenCode Zen Key for paid/authenticated quota
`ROTATING_PROXY_ENABLED`	`false`	Enable dedicated single rotating proxy mode
`ROTATING_PROXY_URL`	`""`	Backconnect rotating proxy URL (`http://user:pass@host:port`)
`ROTATING_PROXY_SESSION_MODE`	`auto`	Session rotation strategy (`auto`, `session_suffix`, `disabled`)
`EGRESS_MODE`	`auto`	Egress mode (`auto`, `proxy_only`, `direct`)
`ALLOW_DIRECT_FALLBACK`	`true`	Fallback to host egress if all proxies are exhausted
`AUTO_PROXY_TARGET`	`100`	Target verified proxy pool size
`MAX_PROXY_POOL`	`150`	Maximum combined proxy pool capacity
`PROXY_QUOTA_FAILOVERS`	`10`	Maximum proxy hops on 429 quota before deadline
`REQUEST_TIMEOUT_MS`	`30000`	Timeout per inference attempt (30 seconds)
`REQUEST_DEADLINE_MS`	`90000`	Total end-to-end inference deadline (90 seconds)
`GLOBAL_CONCURRENCY`	`8`	Maximum simultaneous upstream inferences

License

This project is licensed under the MIT License — see the LICENSE file for details.

Developed for educational and technical research into high-performance AI gateways and reverse-proxy protocol adapters.

Name		Name	Last commit message	Last commit date
client		client
data		data
dist		dist
docs		docs
scripts		scripts
server		server
test		test
.dockerignore		.dockerignore
.env.example		.env.example
.gitattributes		.gitattributes
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
Dockerfile		Dockerfile
LICENSE		LICENSE
README.md		README.md
docker-compose.yml		docker-compose.yml
entrypoint.sh		entrypoint.sh
master.md		master.md
package.json		package.json
test-chat.ps1		test-chat.ps1
test-stream.ps1		test-stream.ps1
test_exhaustive.js		test_exhaustive.js
test_gateway.js		test_gateway.js

Navigation Menu

Repository navigation

About

Repository files navigation

⚡ OpenCode2API v3.2.2

High-Performance OpenAI & Anthropic Compatible Gateway for OpenCode Models

⚠️ Educational & Research Purpose Disclaimer

📖 Table of Contents

What OpenCode2API Does

Architecture & Upstream Quota Reality

Residential ISP vs. Cloud Datacenter IPs

3 Ways to 100% Reliability

Key Features in v3.1.0

Discovered Models Roster

Quickstart: Local Installation (Recommended)

1. Prerequisites

2. Clone & Install

3. Start Gateway

4. Access Gateway & Playground

Quickstart: Docker & Docker Compose

Deploying to Dokploy / Coolify

Zero-Cost Remote Access via Cloudflare Tunnel

Turnkey PowerShell Testing Scripts

1. Test Chat Completions (Any Model)

2. Test Real-Time SSE Token Streaming

API Usage Examples

OpenAI Chat Completions

Anthropic Messages

Streaming in Python

Configuration Reference

License

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

⚡ OpenCode2API v3.2.2

High-Performance OpenAI & Anthropic Compatible Gateway for OpenCode Models

⚠️ Educational & Research Purpose Disclaimer

📖 Table of Contents

What OpenCode2API Does

Architecture & Upstream Quota Reality

Residential ISP vs. Cloud Datacenter IPs

3 Ways to 100% Reliability

Key Features in v3.1.0

Discovered Models Roster

Quickstart: Local Installation (Recommended)

1. Prerequisites

2. Clone & Install

3. Start Gateway

4. Access Gateway & Playground

Quickstart: Docker & Docker Compose

Deploying to Dokploy / Coolify

Zero-Cost Remote Access via Cloudflare Tunnel

Turnkey PowerShell Testing Scripts

1. Test Chat Completions (Any Model)

2. Test Real-Time SSE Token Streaming

API Usage Examples

OpenAI Chat Completions

Anthropic Messages

Streaming in Python

Configuration Reference

License

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages