Feature GuideFebruary 2026

CLI-as-API: Turn Your CLI Subscriptions into REST APIs

Use your existing Claude Code, Codex CLI, or Gemini CLI subscriptions as standard REST APIs. Point any OpenAI, Anthropic, or Google SDK at Polydev and your local CLI handles the request.

What is CLI-as-API?

CLI-as-API is a tunnel that exposes your locally-installed AI CLIs as standard REST API endpoints. If you have a Claude Code, Codex CLI, or Gemini CLI subscription, you can use them as drop-in replacements for the official APIs — no separate API key billing required.

Why? Many developers pay for CLI subscriptions (e.g. Claude Code Max at $200/mo) but also need API access for scripts, IDE integrations, and automation. CLI-as-API bridges this gap — one subscription, both interfaces.

How It Works

The request flow is simple:

Your app sends a standard API request (OpenAI, Anthropic, or Google format) to polydev.ai/api/v1
Polydev validates your pk-* key and checks rate limits
The request is queued and picked up by your local MCP process
Your local CLI (Claude Code, Codex, Gemini) processes the prompt
The response flows back through the tunnel in the original API format

Universal keys: A single pk-* key works with all three providers. Routing is automatic based on the model ID you request.

Getting Started

1. Create an API Key

Go to the CLI-as-API dashboard and click "Create Key". You can optionally set custom rate limits per hour, day, and month.

2. Start the Tunnel

Your Polydev MCP server acts as the tunnel. If you already have Polydev configured in any IDE (Claude Code, Cursor, Cline, etc.), the tunnel starts automatically when the MCP process runs.

3. Make Requests

Point any SDK or HTTP client at https://www.polydev.ai/api/v1 with your pk-* key:

from openai import OpenAI

client = OpenAI(
    base_url="https://www.polydev.ai/api/v1",
    api_key="pk-your-key-here"
)

response = client.chat.completions.create(
    model="claude-opus-4-6",
    messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)

Rate Limits

Each API key has three configurable rate limit windows. You can set custom limits up to your tier maximum when creating a key.

Window	Free	Premium
Per Hour	10	100
Per Day	50	500
Per Month	500	10,000

Rate limits reset at the top of each UTC hour, midnight UTC (daily), and the first of each month. Tunnel requests also count against your monthly message quota (500 messages/month).

Supported Models

Pass any of these model IDs in your API request. The tunnel automatically routes to the correct CLI provider.

Claude (via Claude Code)	GPT (via Codex CLI)	Gemini (via Gemini CLI)
claude-opus-4-6	gpt-5.3-codex	gemini-3-pro-preview
claude-sonnet-4-5-20250929	gpt-5.1-mini	gemini-2.5-flash
claude-haiku-4-5-20251001	o3, o4-mini	gemini-2.5-pro

SDK Examples

OpenAI Python

from openai import OpenAI

client = OpenAI(
    base_url="https://www.polydev.ai/api/v1",
    api_key="pk-your-key-here"
)

# Use any model — routes automatically
response = client.chat.completions.create(
    model="claude-opus-4-6",  # or gpt-5.3-codex, gemini-2.5-pro
    messages=[{"role": "user", "content": "Explain quicksort"}]
)
print(response.choices[0].message.content)

Anthropic Python

import anthropic

client = anthropic.Anthropic(
    base_url="https://www.polydev.ai/api/v1",
    api_key="pk-your-key-here"
)

message = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=4096,
    messages=[{"role": "user", "content": "Explain quicksort"}]
)
print(message.content[0].text)

Google Gemini (REST)

curl https://www.polydev.ai/api/v1/google/generate \
  -X POST \
  -H "x-goog-api-key: pk-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"parts":[{"text":"Explain quicksort"}]}]}'

curl (OpenAI format)

curl https://www.polydev.ai/api/v1/chat/completions \
  -H "Authorization: Bearer pk-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-opus-4-6",
    "messages": [{"role": "user", "content": "Explain quicksort"}]
  }'

IDE Setup

Use CLI-as-API as a custom model provider in any IDE that supports the OpenAI API format.

Cursor

In Cursor Settings > Models > Custom API, add:

Base URL: https://www.polydev.ai/api/v1
API Key:  pk-your-key-here
Model:    claude-opus-4-6

Cline (VS Code)

In Cline settings, select "OpenAI Compatible" as the provider:

API Provider: OpenAI Compatible
Base URL:     https://www.polydev.ai/api/v1
API Key:      pk-your-key-here
Model ID:     claude-opus-4-6

Continue (VS Code)

Add to your ~/.continue/config.json:

{
  "models": [{
    "provider": "openai",
    "title": "Polydev CLI-as-API",
    "model": "claude-opus-4-6",
    "apiBase": "https://www.polydev.ai/api/v1",
    "apiKey": "pk-your-key-here"
  }]
}

Troubleshooting

"Tunnel offline" (503)

Your Polydev MCP process isn't running. Open any IDE with Polydev configured, or check the tunnel status on the dashboard.

"Rate limit exceeded" (429)

You've hit your hourly, daily, or monthly limit. Check which window is exhausted on the dashboard and either wait for reset or adjust your key's limits.

"No matching CLI provider" (503)

The model you requested requires a CLI that isn't connected. For example, requesting a gpt-* model requires Codex CLI to be installed and detected.

"Request timed out" (504)

The CLI took longer than 120 seconds to respond. This can happen with very long prompts or complex requests. Try simplifying the prompt or check that your CLI is responsive.

Ready to get started?

Create your first API key and start using your CLI subscriptions as REST APIs.

Go to CLI-as-API Dashboard More Articles