vm2apiSubscriptions, as a standard API.

A fully isolated VM-level gateway. Official Claude Code and ChatGPT clients run inside their own slots and speak OpenAI and Anthropic protocols.

What is vm2api?

Quick Start

GitHub

Official client, not a fake HTTP stack

Each slot runs the real Claude Code process. Credentials stay on that slot. The gateway does not invent a third-party client fingerprint.

One protocol in, the models you already pay for

Call /v1/messages, /v1/chat/completions, or /v1/responses. Upstream is the subscription behind the slot you imported.

One slot, one egress

Bind a SOCKS5 proxy or the local egress before importing an account. Traffic for that account leaves through that exit.

Zero prompt injection

Identity is aligned at the credential layer. The zero layout does not prepend a fake system persona. official and official_full remain available.

Keys and quotas

The master key can run the panel. Issued sk-vm- keys can only call /v1. Per-slot concurrency and 5h / 7d windows follow the upstream budget.

Console included

Slots, proxy pool, model policy, request logs, and cluster nodes live in the web console on port 8787.

Call

A normal API in front. A real client behind it.

Point any Anthropic or OpenAI compatible client at port 8787. The control plane picks a schedulable slot, the slot's official client talks upstream through that slot's egress, and the caller gets the protocol it sent.

messages.sh

curl -sS http://127.0.0.1:8787/v1/messages \
  -H "Authorization: Bearer $VM2API_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-5",
    "max_tokens": 128000,
    "messages": [{"role":"user","content":"Hello"}]
  }'

Install it on Ubuntu 24.04

Quick Start