Repo public — MIT · 26 commits
LEGION GASPER
Give it a task. It splits that task into pieces, hands each piece to a specialized agent, runs them in parallel, and stitches the results back together. The captain handles the coordination — you don't.
THE MULTI-AGENT AI
LEGIONGASPER is a Python multi-agent system built around one problem: getting multiple LLM agents to work together without you driving every step. You give it a task — it splits the task, recruits specialized agents, runs them in parallel, and merges everything. The repo's own honest version: it's not magic, it's a well-structured Python system that solves a real coordination problem.
It runs on FastAPI, with a live dashboard at localhost:8080 and the API on port 8081 — active agents, task queue, cost per provider, all real-time over WebSocket. The multi-provider router speaks OpenAI, Anthropic and OpenRouter without extra wiring.
Memory works across five layers, from what an agent is doing right now down to a ChromaDB vector store that survives reboots. Redis backs distributed memory across workers. The repo ships an agent factory for spawning agents from templates, plus governance and audit logs.
git clone https://github.com/deathlegion/legiongasper.gitcd legiongasperpip install -r requirements.txtpython -m legiongasper.cli init # API on :8081, dashboard on :8080python -m legiongasper.cli serveWant to help?
LEGIONGASPER needs: Bug reports and PRs on GitHub — the repo's Contributing section covers the setup.. Check CONTRIBUTING.md in the repo, then open a PR — or a bug report with reproduction steps, which is even better.