ccproxy-client (Node)

Zero-dependency client for a ccproxy server. ccproxy exposes an OpenAI- and Anthropic-compatible HTTP API that is answered by the Claude Code CLI. Project home: https://github.com/wubbyweb/ccproxy

What you need

This package is only an HTTP client. Installing it does not install or require Claude Code. You need two things:

1. A running ccproxy server you can reach over HTTP. Claude Code must be installed and logged in on the machine running that server, not on the machine running this client.
2. An API key for that server, created in its admin console (http://<server>:8787/admin).

Two common setups:

| Setup | Who needs Claude Code | Client config | |---|---|---| | Same machine: you run the server locally and your app talks to it | you, on this machine | baseURL default http://127.0.0.1:8787 | | Shared server: someone hosts ccproxy and hands out API keys | only the host machine | baseURL: "https://proxy.example.com" plus the key you were given |

To run a server yourself:

npm install -g ccproxy-server   # needs Node 18+ and a logged-in Claude Code CLI
ccproxy serve
ccproxy admin token             # sign in at http://127.0.0.1:8787/admin and create a key

Install

npm install ccproxy-client

Node 18+ (uses global fetch). Also works in browsers.

Use

import { CCProxy } from 'ccproxy-client';

// apiKey and baseURL also come from CCPROXY_API_KEY / CCPROXY_BASE_URL env vars
const client = new CCProxy({ apiKey: 'sk-ccp-...', baseURL: 'http://127.0.0.1:8787' });

// One-liner
console.log(await client.complete('Explain monads in one sentence.'));

// OpenAI-style
const res = await client.chat.completions.create({
  model: 'sonnet',
  messages: [{ role: 'user', content: 'Hello!' }],
});
console.log(res.choices[0].message.content);

// Streaming
const stream = await client.chat.completions.create({ messages: [{ role: 'user', content: 'Tell me a story' }], stream: true });
for await (const chunk of stream) process.stdout.write(chunk.choices[0]?.delta?.content ?? '');

// Anthropic-style
const msg = await client.messages.create({ max_tokens: 500, messages: [{ role: 'user', content: 'Hi' }] });
console.log(msg.content[0].text);

// Models and health
console.log((await client.models.list()).map((m) => m.id));
console.log(await client.health());

Errors throw CCProxyError with status (HTTP status, 401 for a bad key, 429 for rate limits, 0 for connection failures) and body.

Function calling (tools), response_format, stop sequences and reasoning_effort are passed through; see the project README for what the server supports.

Alternatives

The official openai and @anthropic-ai/sdk packages also work: set baseURL to the server (.../v1 for OpenAI) and use the ccproxy key as the API key.