Knowledge Graph Tools¶

Overview¶

16 unified query tools exposed through three interfaces — CLI, MCP, and LangChain agent — all backed by a single ToolRegistry. Adding a new tool means writing one function and registering it; all interfaces pick it up automatically.

Architecture¶

ToolDefinition(name, function, description, params)
         │
    ToolRegistry (single source of truth)
    ┌──────────────────┐
    │ register()       │
    │ get_tool()       │
    │ get_all_tools()  │
    └──────┬───────────┘
           │
     ┌─────┼──────┐
     ▼     ▼      ▼
    CLI   MCP   LangChain Agent

CLI Usage¶

# Top biomarkers for a disease
uv run python -m src.kg_tools.cli get_top_biomarkers --disease "cardiovascular disease" --limit 5

# Fuzzy entity search
uv run python -m src.kg_tools.cli search_entities --query TP53 --entity_type protein

# Human-readable output
uv run python -m src.kg_tools.cli --human get_protein_interactions --protein TP53

# Raw Cypher (read-only)
uv run python -m src.kg_tools.cli run_cypher_query --query "MATCH (p:Protein) RETURN p.name LIMIT 5"

MCP Server¶

Exposes tools to AI assistants (Claude, VS Code, Kiro) via FastMCP:

{
  "mcpServers": {
    "olink-rag": {
      "command": "uv",
      "args": ["run", "python", "-m", "src.mcp_server.server"],
      "env": {
        "NEO4J_URI": "bolt://localhost:7687",
        "NEO4J_USER": "neo4j",
        "NEO4J_PASSWORD": "your_password"
      }
    }
  }
}

LangChain Agent Interface¶

from src.kg_tools.registry import ToolRegistry
from src.kg_tools.tools import register_all
from src.kg_tools.agent_adapter import get_agent_tools

registry = ToolRegistry()
register_all(registry)
tools = get_agent_tools(registry, db)  # All 16 tools
tools = get_agent_tools(registry, db, groups=["biomarker", "evidence"])  # Filtered

Tool Groups¶

Group	Tools
`biomarker`	get_top_biomarkers, get_disease_proteins, get_evidence_for_association, search_entities
`protein`	get_protein_details, get_protein_interactions, get_protein_pathways, get_shared_diseases
`disease`	get_disease_hierarchy, get_disease_proteins, get_shared_proteins, search_entities
`pathway`	get_protein_pathways, get_pathway_members, search_entities
`evidence`	get_evidence_for_association, get_relationship_evidence_depth, search_entities
`exploration`	search_entities, get_entity_disambiguation, get_graph_statistics, get_community_summary, run_cypher_query

Performance¶

All 16 tool definitions fit in 713 tokens — no need for dynamic tool subsetting. MCP wrapper adds <0.001ms overhead vs direct call. CLI adds ~10ms from argparse.

See KG Tools Guide for full tool reference with schemas, and Benchmark Report for detailed performance data.