Skip to main content
Back

MAESTRO Brings Private Deep-Research Automation to Your own Hardware

An open-source, self-hosted research platform coordinates five AI agents to plan, search, reflect, and write reports over your own document library.

Repo, Full Access
  • MAESTRO is a self-hosted AI research platform with about 1.5k GitHub stars
  • Uses five agents: Controller, Planning, Research, Reflection, Writing in iterative rounds
  • RAG stack built on PostgreSQL + pgvector with dual BGE-M3 embeddings
  • Supports OpenAI, Anthropic, Azure OpenAI with GPT-5, and local OpenAI-compatible LLMs
  • Web search via Tavily, LinkUp, Jina, SearXNG plus a direct arXiv paper fetcher
  • AGPLv3 licensed with Docker deployment, 16GB RAM minimum, GPU recommended

Running deep-research workflows through hosted services like ChatGPT or Perplexity means shipping your queries, and sometimes your documents, off to someone else's servers. MAESTRO is a self-hosted alternative that has quietly grown to around 1.5k GitHub stars. It bundles a multi-agent research loop, a document library with vector search, and a writing assistant into a single Docker stack you can run on your own hardware.

The project describes itself as an AI-powered research platform that manages complex investigations end to end. You define a question, the agents plan the work, pull from your uploaded documents and the web, then produce a cited report. Licensing is AGPLv3, with a commercial option for organizations that cannot comply with AGPL terms.

MAESTRO research report interface

Five agents in a reflection loop

At the core sits a multi-agent system with specialized roles: Controller, Planning, Research, Reflection, and Writing. Rather than firing a single LLM call, the Planning agent breaks a mission into subtasks, Research gathers evidence from documents and web sources, and Reflection critiques the draft to trigger additional rounds. The Writing agent assembles the final report with citations tracked throughout.