MCP Server

Use Abel as a native tool in Claude, GPT, or any MCP-compatible AI assistant.

Installation

terminalbash

npm install -g abel-mcp

# Or use npx (no install)
npx abel-mcp --api-key sk-your-key

Configuration

Add Abel to your MCP settings file:

mcp_settings.jsonjson
{
  "mcpServers": {
    "abel": {
      "command": "npx",
      "args": ["abel-mcp"],
      "env": {
        "ABEL_API_KEY": "sk-your-key"
      }
    }
  }
}

Available Tools

Once connected, your LLM has access to 5 tools:

abel_schema_search

Search Abel's causal world model for variables, communities, and neighborhoods.

abel_causal_explain

Explain what causally drives a variable — parents, children, cross-domain chains.

abel_causal_intervene

Simulate causal intervention: if X changes, what happens to Y?

abel_causal_predict

Predict a variable using its causal Markov blanket.

abel_causal_analyze

Upload data and run full autonomous causal analysis (39 algorithms).

Example Interaction

Once configured, just ask your AI assistant a causal question:

You:

“If oil prices hit $120, what happens to US inflation?”

Claude:

Uses abel_schema_search to find “WTI_Crude” and “CPI”, then calls abel_causal_intervene to compute the causal effect: +2.1% CPI increase via the WTI → CPI causal chain (τ=30h).