BeHive icon

BeHive

by qa10devteam

communityHTTPpypi
View all implementations
120stars|13forks
View on GitHub
Works withClaude DesktopCursorVS CodeWindsurf

About

MCP-native deep research engine for scored claims, entity graphs, and cited reports.

Features

  • Scored atomic claims
  • Entity relationship graphs
  • Structured JSON outputs
  • Synthesized reports with citations
  • MCP endpoint and REST API
  • Bring-your-own-key model routing

Use Cases

Research agent builder

Use BeHive to turn research questions into cited, machine-readable claims and graphs.

Self-hosted AI team

Run a local research service with PostgreSQL storage and MCP integration.

What This Server Can Do

MCP servers expose three types of capabilities to AI clients. Here's what BeHive supports.

Tools

Supported3 tools

Functions your AI client can call to perform actions — like querying a database, creating a file, or calling an API.

How to use: Tools run automatically when your AI client decides they're needed. Ask your AI assistant to perform a task, and it will invoke the right tool.

Resources

Supported

Read-only data sources this server exposes — like files, database schemas, or API responses your AI client can read for context.

How to use: Resources are loaded as context when your AI client needs background information. Ask about the data this server manages, and resources get pulled in automatically.

Prompts

Not available

This server does not provide pre-built prompt templates.

For the full list of available tools, resources, and prompts, check the README on GitHub.

Available Tools (3)

research_mission

Run a structured deep-research mission and return claims, entities, and report data.

claim_search

Search stored claims and research memory.

entity_graph

Inspect entities and relationships extracted during research.

Frequently Asked Questions