AI workforce orchestrator. One dashboard, any provider. No vendor lock-in.
Route tasks to Claude, Codex, Ollama, Aider, or any CLI tool from a single dashboard. Workers run on your machines with your API keys. We orchestrate. We don't eavesdrop.
Website | Quickstart | Changelog
# 1. Install and run the server
pip install -r requirements.txt
python app.py
# 2. Connect a worker (any machine)
export MODELREINS_URL=http://your-server:8484
export MODELREINS_TOKEN=your-token
export MODELREINS_WORKER=my-worker
npx @mediagato/modelreins-workerOr use the hosted version at app.modelreins.com.
- Multi-provider routing — Claude for architecture, Codex for implementation, Ollama for free
- Direct API keys — workers use your keys on your machines, not subscriptions that can be revoked
- Conversation context — follow-up messages include full history
- Preflight validation — workers verify their environment before claiming jobs
- Structured error classification — spawn_failed, auth_failed, timeout, rate_limited
- 3-level killswitch — kill worker, kill tenant, global emergency stop (all plans)
- Prompt sanitization — strips hidden HTML, invisible unicode, injection attempts
- Engine brain — shared product knowledge that helps workers make better decisions
- Signed audit trail — every action HMAC-signed and logged
| Provider | Type | Setup |
|---|---|---|
| Claude Code | CLI | MODELREINS_PROVIDER=claude |
| Codex | CLI | MODELREINS_PROVIDER=codex |
| Aider | CLI | MODELREINS_PROVIDER=aider |
| Ollama | API | MODELREINS_PROVIDER=ollama |
| Any CLI tool | Custom | MODELREINS_PROVIDER=custom MODELREINS_COMMAND=your-tool |
Add new providers via YAML config in providers/ — no code changes needed.
YOU --> MODEL ROUTER --> Workers (your machines)
| |
Brain (server) AI Tools (your API keys)
|
Engine Knowledge
- Server: Python + FastAPI + SQLite. One process, one file database.
- Workers: Node.js daemon. Polls for jobs, spawns AI tools, streams output.
- Dashboard: Real-time via WebSocket. Job feed, worker status, dispatch bar.
- Brain: Multi-tenant state — engine (shared), tenant (isolated), ops (admin).
| Self-Hosted | SaaS | |
|---|---|---|
| Server | Your machine | app.modelreins.com |
| Data | Your database | Tenant-isolated |
| Workers | Unlimited | 2 (Free) / 10 (Pro) / 50 (Team) |
| Jobs | Unlimited | Unlimited |
| Cost | Free | Free / $29 / $79 |
- Python 3.12+ / FastAPI / SQLite
- Node.js 18+ (worker daemon)
- React 19 / Vite / TypeScript / Tailwind (dashboard v2)
- Vanilla JS (dashboard v1, production)
Business Source License 1.1 (BSL) — free to self-host, cannot be offered as a competing hosted service. See LICENSE for details.
Built by MEDiAGATO