Advertise
MCP.so
Sign In

AutoGen-Compatible Multi-Agent Research POC with Ollama + BraveS…

@chin3

Repository

Basic information

Category

AI & Agents

Transports

stdio

Author

chin3

About AutoGen-Compatible Multi-Agent Research POC with Ollama + BraveSearch MCP Server

This project is a proof of concept for running a local-first multi-agent system using: 🤖 Local LLMs via Ollama 🧩 Simple function/tool-call detection using <tool_call>... 🔍 Brave Search API or optional Brave MCP plugin server 🧠 Two collaborating agents: Searcher and Synthesize

#multi-agent#ollama#brave-search

AutoGen-Compatible Multi-Agent Research POC with Ollama + BraveSearch MCP Server config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "Multi-Agent-Research-POC": {
      "command": "python",
      "args": [
        "main.py"
      ]
    }
  }
}

AutoGen-Compatible Multi-Agent Research POC with Ollama + BraveSearch MCP Server tools

Tool list not available yet

This listing does not currently have a tool list to display. This does not mean the server has no tools; check its documentation for supported capabilities.

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is 🧠 AutoGen-Compatible Multi-Agent Research POC with Ollama + BraveSearch?

A proof-of-concept for a local-first multi-agent system using Ollama (local LLMs) and Brave Search. It features two collaborating agents—Searcher and Synthesizer—that detect tool calls via <tool_call> syntax and can optionally integrate with the Brave MCP plugin server. Built for developers exploring autonomous, tool-using agents.

How to use 🧠 AutoGen-Compatible Multi-Agent Research POC with Ollama + BraveSearch?

Clone the repo, install Python dependencies (pip install -r requirements.txt), set a BRAVE_API_KEY in .env, run Ollama locally (ollama run llama3:8b), then execute python main.py. To switch from the default Brave Search API to the Brave MCP plugin, start the plugin server (npx @modelcontextprotocol/server-brave-search) and update tools/tool_registry.py.

Key features of 🧠 AutoGen-Compatible Multi-Agent Research POC with Ollama + BraveSearch

  • Local-first multi-agent system with Searcher and Synthesizer agents
  • Web search via Brave Search API or Brave MCP plugin server
  • Tool-call detection using <tool_call> syntax
  • Supports switching between API and MCP backends
  • Designed for the Microsoft AI Agents Hackathon

Use cases of 🧠 AutoGen-Compatible Multi-Agent Research POC with Ollama + BraveSearch

  • Conduct web research entirely with local AI agents
  • Synthesize multiple search results into a coherent summary
  • Prototype autonomous, tool-using agents without cloud dependencies
  • Test multi-agent collaboration patterns with local LLMs

Frequently asked questions

1

What does this project do?

It runs two agents (Searcher and Synthesizer) locally with Ollama, queries the web via Brave Search (API or MCP plugin), and produces a final summary from search results.

2

What runtime dependencies are required?

Ollama (with a model like `llama3:8b`), Python 3, and a Brave Search API key. Optionally, Node.js/npx for the MCP plugin.

3

How do I switch between the Brave API and the MCP plugin?

By default the tool uses `call_brave_api`. To use the MCP plugin, start the plugin server (`npx @modelcontextprotocol/server-brave-search`) and change `tools/tool_registry.py` to use `call_brave_mcp_server` instead.

4

Where does data from searches live?

Search results are fetched from Brave and processed entirely locally; no data is stored externally. The project saves no session logs by default.

5

Is this project ready for production?

No—it is a proof of concept built for the Microsoft AI Agents Hackathon. The README lists planned improvements like a Planner agent, more tools, a UI, and API wrapping.

Comments

More AI & Agents MCP servers

HeyLead: LinkedIn Outreach & Lead Generation MCP Server

D4umak

HeyLead: AI agent for LinkedIn outreach. Finds the right people, writes in your voice, follows up and handles replies, from Claude Code, Cursor, any MCP client or a web dashboard.

9

xhostd

xhostd

Agent-first hosting. From Claude, Cursor, or Codex, your agent can create an app, push code, and deploy it to a live HTTPS URL, with managed Postgres, custom domains, environment variables, backups, and rollback.

GoBuy - Product Trust

gobuy-ai

GoBuy gives AI agents marketplace-evidence trust scores. One call returns a 0–100 Evidence Score with observed/missing signals, freshness, and product context. Tools: • check_product_trust — inspect one Amazon, Walmart,

Lumify Sports Intelligence

Lumify

Agent-ready sports intelligence API: live scores, odds, line movement, public betting splits, and explainable bet confidence — via 16 MCP tools at https://lumify.ai/mcp. Get a free key instantly — no signup, email, or ca

q-ring

I4cTime

Quantum-inspired keyring for AI coding agents. Secure secrets with superposition, entanglement, tunneling, and teleportation — CLI + 44 MCP tools.

3
Advertise