MCP Server Overview

The Hubify MCP server exposes labs, experiments, agents, and knowledge to AI coding assistants.

MCP Server

Hubify Labs includes a Model Context Protocol (MCP) server that exposes your lab's data and capabilities to AI coding assistants like Claude Code, Cursor, and other MCP-compatible tools.

What It Does

The MCP server lets AI assistants:

  • Read lab status, experiment results, agent activity, and knowledge base entries
  • Execute experiments, create tasks, trigger reviews, and manage pods
  • Access structured prompts for research workflows

Instead of copy-pasting data between your lab and your coding environment, the MCP server provides direct, structured access.

Architecture

┌──────────────────┐    MCP Protocol    ┌──────────────────┐
│  Claude Code     │◄──────────────────►│  Hubify MCP      │
│  Cursor          │    (stdio/SSE)     │  Server           │
│  Other MCP Host  │                    │                   │
└──────────────────┘                    └────────┬──────────┘
                                                 │
                                        ┌────────▼──────────┐
                                        │  Hubify Labs API  │
                                        │  api.hubify.com   │
                                        └───────────────────┘

The MCP server runs locally and communicates with the Hubify API on your behalf. It translates MCP tool calls and resource reads into API requests.

Capabilities

MCP FeatureHubify Implementation
Tools18 tools for experiments, agents, tasks, pods, papers, knowledge
ResourcesLab status, experiment data, knowledge entries, paper drafts
PromptsPre-built prompts for research analysis, paper writing, code review

Quick Start

# Install the MCP server
npm install -g @hubify/mcp-server

# Configure with your API key
hubify mcp setup --api-key $HUBIFY_API_KEY

# Test the connection
hubify mcp test

Then add the server to your MCP host configuration. See Setup for detailed instructions.

When to Use

Use the MCP server when you want your AI assistant to:

  • Query experiment results without leaving your editor
  • Create and monitor experiments from within Claude Code
  • Search the knowledge base during paper writing
  • Check pod status and cost while debugging on a pod
  • Trigger standups and reviews from your development environment

Next Steps

  • Setup — Install and configure the MCP server

  • Tools — Browse available MCP tools

  • Resources — Browse available MCP resources

  • Prompts — Browse pre-built research prompts

← Back to docs index