agentoracle-mcp

Add trust verification to Claude, Cursor, Windsurf, or any MCP client in 60 seconds.

AgentOracle verifies claims before your agent acts on them. Per-claim confidence scores, ACT/VERIFY/REJECT recommendations, and 4-source verification — available as an MCP tool with one command.

Install

npx agentoracle-mcp

That's it. No API keys. No accounts. No wallet setup required to start.

Add to Claude Desktop

Open your Claude Desktop config file:

Mac: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

Add this:

{
  "mcpServers": {
    "agentoracle": {
      "command": "npx",
      "args": ["-y", "agentoracle-mcp"]
    }
  }
}

Restart Claude Desktop. You now have trust verification as a tool.

Add to Cursor

Open Cursor Settings → MCP → Add Server:

{
  "agentoracle": {
    "command": "npx",
    "args": ["-y", "agentoracle-mcp"]
  }
}

Add to Windsurf

Open Windsurf Settings → MCP Servers → Add:

{
  "agentoracle": {
    "command": "npx",
    "args": ["-y", "agentoracle-mcp"]
  }
}

What you can do once installed

Ask Claude, Cursor, or any MCP client:

• "Verify these claims before I include them in my report"

• "Is it true that OpenAI acquired Anthropic in 2026?"

• "Research AI agent frameworks and tell me what confidence score each claim gets"

• "Fact-check this paragraph before I publish it"

AgentOracle breaks the input into individual claims, runs each one through 4 independent sources, and returns a confidence score and verdict for every single claim.

What comes back

{
  "overall_confidence": 0.91,
  "recommendation": "act",
  "claims": [
    {
      "claim": "LangGraph leads agent frameworks in 2026",
      "verdict": "supported",
      "confidence": 0.94,
      "evidence": "Confirmed across 4 independent sources"
    },
    {
      "claim": "OpenAI acquired Anthropic in early 2026",
      "verdict": "refuted",
      "confidence": 0.04,
      "correction": "Anthropic remains independent as of April 2026"
    }
  ]
}

How verification works

Every claim runs through 4 sources in parallel:

1. Sonar — real-time web research

2. Sonar Pro — deep multi-step analysis

3. Adversarial — actively tries to disprove the claim

4. Gemma 4 — claim decomposition and confidence calibration

Consensus builds the score. Contradiction flags the risk.

Pricing

Payments via x402 protocol — USDC on Base, SKALE (gasless), or Stellar. No subscriptions. No minimums.

The free /preview endpoint works with no wallet or setup. Use it to try before you pay.

Try it now without installing anything

curl -X POST https://agentoracle.co/preview \
  -H "Content-Type: application/json" \
  -d '{"query": "OpenAI acquired Anthropic in 2026"}'

Related

• agentoracle.co — main site + live demo

• Trust Layer docs — full API reference

• langchain-agentoracle — LangChain integration

• crewai-agentoracle — CrewAI integration

• x402 manifest — agent-native pricing discovery

Built by TK Collective · x402 native · Base · SKALE · Stellar