Model Context Protocol

Connect your agent to Keppo.

One MCP endpoint, two tools, and answers backed by scoped graph evidence. Configure once and your agent can ask Keppo from inside its tool loop.

MCP endpoint

https://mcp.keppo.ai

POST JSON-RPC to this URL. streamable-http transport.

Step 1

Add Keppo to your client

Pick the agent you use. The config snippet below registers Keppo as an MCP server.

Claude Code

One-line CLI install.

claude mcp add keppo https://mcp.keppo.ai

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "keppo": {
      "transport": "streamable-http",
      "url": "https://mcp.keppo.ai"
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project (or globally).

{
  "mcpServers": {
    "keppo": {
      "url": "https://mcp.keppo.ai"
    }
  }
}

Step 2

Once registered, your agent gets these tools in its loop.

ask_public_dev_knowledge

Ask Keppo's public developer knowledge graph for a provenance-stamped answer.

query

report_feedback

Report incorrect, stale, unsafe, or missing-context feedback for a specific answer.

questionfeedback_text

Step 3

Quick smoke check that your endpoint speaks JSON-RPC.

curl -sS https://mcp.keppo.ai \
  -H "content-type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | jq

Background

Model Context Protocol is the open standard agents use to discover and call tools — like USB-C for LLMs.

Verified, not hallucinated

Keppo grades answers by evidence strength, claim coverage, and faithfulness before returning them to your agent.

One source of truth

Claims, evidence refs, answer citations, and freshness checks live in the graph so web and MCP clients see the same context.