What is this?
Agent2API is a local gateway that takes the logged-in sessions of AI desktop clients like WorkBuddy, CatPaw and AutoClaw and exposes them through one shared OpenAI-compatible API, so any client with a custom baseurl can reuse those accounts' model quota without an upstream API key.
Agent2API is a local gateway that takes the logged-in sessions of AI desktop clients like WorkBuddy, CatPaw and AutoClaw and exposes them through one shared OpenAI-compatible API, so any client with a custom baseurl can reuse those accounts' model quota without an upstream API key.Is this a desktop app, a server, or both?It ships in three forms: a desktop app with a tray icon (Windows/macOS/Linux), a headless server that registers as a system service (Windows/Linux), and a Docker image (amd64/arm64) where the admin panel and the gateway share the same port.
It ships in three forms: a desktop app with a tray icon (Windows/macOS/Linux), a headless server that registers as a system service (Windows/Linux), and a Docker image (amd64/arm64) where the admin panel and the gateway share the same port.How do I get it running?Pick one of the three builds (desktop installer, headless server package, or Docker image), start it, open the web panel to register an admin account, create a gateway Key, then add a provider account and point your client's baseurl at the local endpoint.
Pick one of the three builds (desktop installer, headless server package, or Docker image), start it, open the web panel to register an admin account, create a gateway Key, then add a provider account and point your client's baseurl at the local endpoint.Do I need Node.js or any extra runtime to run it?No — the released desktop, server, and Docker builds don't require Node; Node.js >=18.17 is only needed if you're building the frontend yourself.
No — the released desktop, server, and Docker builds don't require Node; Node.js >=18.17 is only needed if you're building the frontend yourself.Is using this against the providers' terms a risk?Yes — it's an unofficial client that reuses your own login sessions, which may violate upstream terms of service; risks like rate-limiting, flagging, or account bans are entirely on the user, and commercial or quota-bypassing use is explicitly banned.
Yes — it's an unofficial client that reuses your own login sessions, which may violate upstream terms of service; risks like rate-limiting, flagging, or account bans are entirely on the user, and commercial or quota-bypassing use is explicitly banned.Who is this realistically for?It's aimed at individual users of specific Chinese AI coding/chat clients (WorkBuddy, CatPaw, AutoClaw, Qoder, Trae, etc.) who want to consolidate their own already-logged-in accounts behind one OpenAI-style endpoint for personal, non-commercial use.
It's aimed at individual users of specific Chinese AI coding/chat clients (WorkBuddy, CatPaw, AutoClaw, Qoder, Trae, etc.) who want to consolidate their own already-logged-in accounts behind one OpenAI-style endpoint for personal, non-commercial use.Which platforms does it actually run on?Desktop builds cover Windows, macOS and Linux; the headless server build covers Windows Server and Linux; the Docker image supports both amd64 and arm64.
Desktop builds cover Windows, macOS and Linux; the headless server build covers Windows Server and Linux; the Docker image supports both amd64 and arm64.How many providers does it support out of the box?17 built-in providers including WorkBuddy, 小浣熊 (CatPaw), Qoder and Trae, plus support for adding custom providers.
17 built-in providers including WorkBuddy, 小浣熊 (CatPaw), Qoder and Trae, plus support for adding custom providers.What happens when an account hits a rate limit?Accounts are queued by priority and the gateway automatically rotates to another account on a 429 response, with a cooldown period so it doesn't keep hammering a limited account.
Accounts are queued by priority and the gateway automatically rotates to another account on a 429 response, with a cooldown period so it doesn't keep hammering a limited account.Does it refresh expired login tokens automatically?Yes — credentials auto-renew before expiry, near-expiry tokens can be bulk-refreshed, and the model catalog is kept from going stale as part of the same scheduled maintenance.
Yes — credentials auto-renew before expiry, near-expiry tokens can be bulk-refreshed, and the model catalog is kept from going stale as part of the same scheduled maintenance.What kind of analytics does the dashboard show?Request volume, success rate, token consumption, top models/accounts, a usage donut chart, a 365-day activity heatmap, and a 24-hour cache-hit-rate/token-consumption trend chart.
Request volume, success rate, token consumption, top models/accounts, a usage donut chart, a 365-day activity heatmap, and a 24-hour cache-hit-rate/token-consumption trend chart.What are similar projects?workbuddy2api-panel, casbin-gateway, vm2api, ThinkWatch-Lite, nofwl, fanzha-ai-proxy, ccx, workbuddy2api-hub
workbuddy2api-panel, casbin-gateway, vm2api, ThinkWatch-Lite, nofwl, fanzha-ai-proxy, ccx, workbuddy2api-hubWhich topics is it in?gateway, openai-api, arm64, reverse-proxy, tauri, macos-app, windows-app, openai-compatible, proxy, dashboard, rust, gui
gateway, openai-api, arm64, reverse-proxy, tauri, macos-app, windows-app, openai-compatible, proxy, dashboard, rust, gui