acp2openai

OpenAI-compatible HTTP facade over ACP (Kilo Code + opencode). Point any OpenAI SDK at it, get Kilo/opencode routing transparently.

Quickstart

# 1. Start ACP backend (one or both)
kilo serve --port 4780
opencode serve --port 4790

# 2. Start acp2openai
npx acp2openai                  # defaults: :4800, kilo :4780, opencode :4790
npx acp2openai --port 8080      # custom port

# 3. Point any OpenAI SDK at http://localhost:4800/v1
#    Or open http://localhost:4800/ for the built-in live-streaming demo

Demo

• Built-in: http://localhost:4800/ when acp2openai is running (no CORS, full streaming)
• GH Pages: https://anentrypoint.github.io/acp2openai (browser-blocked from loopback by Chrome PNA — use built-in demo instead)

Usage

import OpenAI from 'openai';
const client = new OpenAI({ baseURL: 'http://localhost:4800/v1', apiKey: 'none' });

const stream = await client.chat.completions.create({
  model: 'kilo/x-ai/grok-code-fast-1:optimized:free',
  messages: [{ role: 'user', content: 'hi' }],
  stream: true,
});
for await (const chunk of stream) process.stdout.write(chunk.choices[0]?.delta?.content || '');

Model selection

Model IDs: <backend>/<model> where <backend> ∈ {kilo, opencode}.

Examples:

• kilo/x-ai/grok-code-fast-1:optimized:free (free, default)
• kilo/kilo-auto/free
• opencode/minimax-m2.5-free

Bare model IDs (no prefix) route to kilo.

Endpoints

• POST /v1/chat/completions — streaming + non-streaming
• GET /v1/models — live-probes both backends
• GET /health

Why

ACP (Kilo/opencode HTTP serve mode) has a custom protocol: POST /session, POST /session/:id/message, SSE /event. Using it from standard OpenAI-SDK code requires rewriting clients. acp2openai maps:

• OpenAI messages[] → ACP text part
• ACP message.part.delta → OpenAI delta.content / delta.reasoning_content (tracked via partID → part.type lookup)
• ACP tool part state → OpenAI tool_calls
• ACP session.idle / session.turn.close → finish_reason: stop

Streaming uses ACP's true-delta SSE (not accumulating snapshots) for minimum per-token latency.

Config

CLI flags or env:

• --port / PORT (4800)
• --kilo <url> / ACP_KILO_URL (http://localhost:4780)
• --opencode <url> / ACP_OPENCODE_URL (http://localhost:4790)

License

MIT