GEROLAMO MCP — QUICK SETUP

Run: npx gerolamo-mcp setup

Or add manually to ~/.claude/mcp.json:

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

Get your API key at gerolamo.org → Connect.

WHAT YOUR AGENT CAN DO

Scout — Semantic search across 36,000+ GitHub repos, arXiv papers, and HuggingFace models. Every entity scored 1-10 for defensibility with frontier-lab obsolescence risk assessment.

Compose — Fuse multiple technologies into architecture specs, research briefs, or comparison reports. Generate SPEC.md files with project structure, commands, and integration plans.

Analyze — Score dependency stacks for weakest-link risk. Profile creator authority and collaboration networks. Detect defensible clusters in any technology domain.

Optimize — Get cost-optimized foundation model recommendations based on task requirements, capability needs, and budget. Compare pricing across 9 providers with benchmark data.

Invent — Propose speculative "meta molecules" by fusing existing technologies into hypothetical new capabilities. Your agent defines what should exist, links it to parent primitives for lineage tracking, and shares it with the network. When someone builds it for real, connect it to the live URL and it enters the intelligence corpus as a realized entity. Think of it as filing a patent on a technology combination before it exists.

29 TOOLS

Intelligence Search: query_intelligence, search_intelligence, find_sleepers, find_alternatives

Analysis: score_stack, explain_score, analyze_competitive_landscape, explore_connections

Composition: compose_molecules, save_composition, suggest_tools

Briefs: get_intelligence_brief, get_my_latest_intelligence

Creators: get_creator_profile, get_creator_network, get_creator_authority, find_defensible_clusters

Foundation Models: recommend_model, check_model_pricing, compare_foundation_models, get_domination_risk

Workspace: create_workspace, add_to_workspace, submit_molecule Topics: get_tracked_topics

Lineage: submit_meta_molecule, realize_meta_molecule, trace_lineage, find_family

EXAMPLE WORKFLOWS

Build something new from proven primitives:

suggest_tools("autonomous drone navigation system")
→ find_sleepers(query="SLAM flight controller", min_score=6)
→ compose_molecules(entity_ids=[...], mode="compose")
→ save_composition(workspace_name="Drone Stack", mode="compose", result=...)
Pick the right model for your project:
recommend_model(task="vision-based document extraction", require="vision,structured_output", prefer="value")

Threat-check before you ship:

score_stack(entity_ids=["dep1", "dep2", "dep3"])
→ find_alternatives(entity_id="weakest-link")
Propose a new technology that should exist:
query_intelligence(question="real-time SLAM for drones")
→ submit_meta_molecule(title="Edge SLAM Fusion", description="...", parent_entity_ids=["slam-lib", "drone-fw"])
→ trace_lineage(entity_id="new-meta-id", direction="ancestors")

LINKS

• gerolamo.org — Web interface
• npm package — gerolamo-mcp
• llms.txt — Agent discovery file