Gerolamo MCP

Connect your agent to Gerolamo — competitive intelligence across 36,000+ open-source projects, research papers, and ML models, plus foundation model economics via Via Mentis.

Every entity is scored for defensibility (1-10), frontier-lab risk, threat profile, and composability. Your agent can search, analyze, compare, compose technology stacks, and get model recommendations.

Quick Setup

npx gerolamo-mcp setup

This adds Gerolamo to your Claude Code MCP configuration. You'll need an API key:

1. Go to gerolamo.org and request an account
2. Once approved, sign in
3. Go to Connect and generate an API key

Manual Setup

Add to ~/.claude/mcp.json:

{
  "mcpServers": {
    "gerolamo": {
      "url": "https://gerolamo.onrender.com/mcp/sse",
      "headers": {
        "X-API-Key": "glm_your_key_here"
      }
    }
  }
}

Programmatic Usage

import { getConfig, TOOLS } from "gerolamo-mcp";

// Generate MCP config
const config = getConfig("glm_your_key_here");

// List available tools
console.log(Object.keys(TOOLS)); // 31 tools

Available Tools (31)

Intelligence Search

• query_intelligence — Semantic search across all corpora
• search_intelligence — RAG-synthesized answer to a research question
• find_sleepers — High-defensibility, low-traction hidden gems
• find_alternatives — Find projects that could replace a given entity

Analysis

• score_stack — Weakest-link defensibility analysis for a dependency set
• explain_score — Full reasoning behind an entity's defensibility score
• analyze_competitive_landscape — Velocity-ranked topic analysis
• explore_connections — 5 rings of intelligence context around an entity

Composition

• compose_molecules — Fuse entities into specs, comparisons, or research briefs
• save_composition — Persist a generated composition with lineage tracking and shareable URL
• suggest_tools — Recommend which Gerolamo tools to use for a task

Intelligence Briefs

• get_intelligence_brief — Combined situation report: sleepers, trending, breakouts
• get_my_latest_intelligence — Latest subscription alert results

Creator Analysis

• get_creator_profile — Creator portfolio with authority stats
• get_creator_network — Creator collaboration graph
• get_creator_authority — Creator authority score lookup
• find_defensible_clusters — Cluster detection across the corpus

Foundation Models (Via Mentis)

• recommend_model — Recommend the best model for a task based on capabilities, quality, and cost
• check_model_pricing — Foundation model pricing lookup
• compare_foundation_models — Side-by-side model comparison
• get_domination_risk — AI capability domination risk analysis

Topic Intelligence

• get_tracked_topics — List all tracked topics with entity counts, defensibility, and risk stats

Meta Molecules & Lineage

• submit_meta_molecule — Create a speculative meta molecule with required parent lineage
• realize_meta_molecule — Connect a meta molecule to a real URL and queue for ingestion
• trace_lineage — Trace ancestors or descendants of any entity or meta molecule
• find_family — Full lineage family — ancestors, descendants, and direct edges

Workspace

• create_workspace — Create a named workspace with entities
• add_to_workspace — Add entities to an existing workspace
• submit_molecule — Submit a URL for ingestion into the corpus

Workflows

• list_workflows — List all available agent workflow templates
• get_workflow — Get a workflow's full prompt template by slug

Agent Workflows

Workflows are multi-step prompt templates that chain MCP tools into complete intelligence operations. Instead of figuring out which tools to call and in what order, your agent loads a workflow and follows the steps.

"Run the Domain Discovery workflow for autonomous agents"
→ agent calls get_workflow("domain-discovery")
→ agent reads the 6-step prompt
→ agent executes each step using MCP tools, thinking between steps
→ you end up with: meta molecules, compositions, submitted research

8 workflows available:

| Workflow | What it does | Difficulty | |----------|-------------|------------| | domain-discovery | Deep-dive a domain, find gaps, synthesize new ideas, build spec | Intermediate | | cross-domain-intersection | Find novel overlaps between tracked domains | Advanced | | sleeper-hunt | Find hidden gems and build something from them | Intermediate | | stack-audit | Audit dependencies, find weak links, discover alternatives | Beginner | | creator-deep-dive | Profile a creator, map network, predict next move | Intermediate | | investment-thesis | Market dynamics, defensible players, buy/watch/pass | Advanced | | model-selection | Compare foundation models for a use case | Beginner | | trend-spotter | Spot emerging trends, 90-day predictions | Intermediate |

Browse and contribute workflows at gerolamo-workflows.

Example Tool Chains

Scout before you build:

suggest_tools("I need to build an autonomous drone system")
→ get_intelligence_brief(topic="autonomous drone middleware")
→ find_sleepers(query="flight controller SLAM", min_score=6)
→ compose_molecules(entity_ids=[...], mode="compose")
→ save_composition(workspace_name="Drone Stack", mode="compose", result=<output>, entity_ids=[...])

Pick the right model for your project:

recommend_model(task="vision-based document extraction", require="vision,structured_output", prefer="value")
→ check_model_pricing(provider="google")
→ compare_foundation_models(model_names="Gemini Flash, GPT-4o Mini, Claude Haiku")

Threat-check your dependencies:

score_stack(entity_ids=["uuid1", "uuid2", "uuid3"])
→ explain_score(entity_id="weakest-link-uuid")
→ find_alternatives(entity_id="weakest-link-uuid")

Compare before you choose:

query_intelligence(question="vector database")
→ compose_molecules(entity_ids=[...], mode="compare")

CLI Commands

npx gerolamo-mcp setup          # Configure Claude Code
npx gerolamo-mcp setup --key K   # Configure with specific key
npx gerolamo-mcp info             # List all tools

Links

• Gerolamo — Web interface
• gerolamo-workflows — Agent workflow templates (open source)
• API Docs — OpenAPI documentation
• llms.txt — Agent discovery file