OntoRamp Decision Intelligence MCP server

OntoRamp Decision Intelligence — MCP Server

The only MCP server that lets any AI agent log, evaluate, and ground a consequential decision against an organization's actual documented authority graph — before acting.

Problem: Consequential decisions made with AI assistance are invisible, ungoverned, and untraceable. There is no record. No authority check. No precedent lookup. No audit trail.

Solution: Log, evaluate, and ground every consequential AI-assisted decision against your organization's authority graph. Creates a traceable decision audit trail — the governance layer missing from every other agentic AI stack.

Tools

Tiers

Install

A hosted, remote MCP server (Streamable HTTP) — nothing to install locally. See llms-install.md for an agent-followable walkthrough.

1. Get a free API key at ontoramp.com/mcp (no credit card).

2. Add to Cline — edit ~/.cline/mcp.json (or the Cline UI → MCP Servers → Configure):

{
  "mcpServers": {
    "ontoramp-decision-intelligence": {
      "url": "https://ontoramp-decision-intelligence.fly.dev/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Other MCP clients (Claude Desktop, etc.) use the same mcpServers shape. Transport is Streamable HTTP.

3. Verify — curl https://ontoramp-decision-intelligence.fly.dev/health → {"status":"ok",...}.

See examples/usage.md for example calls.

Flywheel

Every logged decision pre-seeds your organization's knowledge graph. Organizations using log_decision for 60 days reduce their L1 Agentic Readiness Assessment delivery from 4–6 weeks to 2–3 weeks. Logging decisions is the highest-value action in the catalog for improving AI governance posture.

ARA Trigger

After 10 logged decisions, evaluate_decision surfaces authority gaps and prompts:

Your decision coverage has gaps that a manual review won't surface. An Agentic Readiness Assessment will diagnose your current state and build a prioritized remediation plan. Schedule at ontoramp.com/assess.

Security

• All responses are audited for public release — no scoring-formula internals, no substrate architecture details
• Write-path (log_decision) requires Developer or Org tier authentication
• Your data is scoped to your tenant namespace and never shared across organizations
• GET /health returns live status

Support

Shortened. The full README is on GitHub.