gpt-researcher

name	gpt-researcher
description	Autonomous multi-source research agent with MCP server. Planner/executor architecture, 5 MCP tools. Requires 3 API keys (Tavily + Anthropic + embeddings). Research-flavor.
auto-activate	false

gpt-researcher

Autonomous multi-source research agent with a planner/executor architecture. Gathers information from multiple sources in parallel.

Upstream: https://github.com/assafelovic/gpt-researcher MCP Server: https://github.com/assafelovic/gptr-mcp

MCP Server Setup

Add to your project's .mcp.json:

{
  "gpt-researcher": {
    "type": "stdio",
    "command": "uvx",
    "args": ["gptr-mcp"],
    "env": {
      "TAVILY_API_KEY": "<your-key>",
      "ANTHROPIC_API_KEY": "<your-key>",
      "FAST_LLM": "anthropic:<your-fast-model>",
      "SMART_LLM": "anthropic:<your-smart-model>"
    }
  }
}

Replace <your-fast-model> and <your-smart-model> with your preferred Claude model IDs before use.

MCP Tools

Tool	Purpose
deep_research	Full multi-source research on a topic
quick_search	Fast single-source lookup
write_report	Generate structured report from research
get_research_sources	List sources used in research
get_research_context	Retrieve accumulated research context

API Keys Required

1. TAVILY_API_KEY — search engine (required)
2. ANTHROPIC_API_KEY — LLM provider (required)
3. Embedding provider key — for semantic search (required)

How It Complements Loam

• /researcher — quick single-pass web research
• GPT-Researcher — autonomous multi-source deep research with structured reports

Requirements

• Python 3.11+
• Three API keys minimum
• MCP server is from separate repo (gptr-mcp), early-stage

Notes

• Apache-2.0 license
• Originally OpenAI-focused, but has native Anthropic provider support
• MCP server is early-stage (separate repo) — expect rough edges
• Per-invocation cost varies by research depth