Model Context Protocol

Use Philidor in Claude and ChatGPT

Connect the Philidor MCP server and your AI assistant can answer DeFi risk questions with independent, sourced vault risk scores instead of guessing. Free, no API key, no account.

10 toolsStreamable HTTPNo authentication

https://mcp.philidor.io/api/mcp

What you can ask

Once connected, these are answered from live Philidor data rather than model recall.

What are the safest USDC vaults on Ethereum?
Compare the Gauntlet and Steakhouse USDC vaults on Morpho.
What is the risk breakdown for this vault, and why is it not Prime?
Give me a market overview of DeFi vault risk right now.

Setup

Claude Code

One command:

claude mcp add philidor --transport http https://mcp.philidor.io/api/mcp

Claude Desktop

Open Settings, then Connectors, then add a custom connector with the server URL above. Desktop builds that still require local config can use an mcp-remote bridge:

{
  "mcpServers": {
    "philidor": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.philidor.io/api/mcp"]
    }
  }
}

Restart Claude Desktop after saving.

ChatGPT

Custom MCP connectors require Developer Mode. Open Settings, then Apps & Connectors, open Advanced settings, and enable Developer Mode. Then add a connector pointing at the server URL, with transport Streamable HTTP and authentication set to None.

https://mcp.philidor.io/api/mcp

Developer Mode is available on Plus, Pro, Business, Enterprise, and Education plans on the web app. On Business and Enterprise workspaces, an admin must enable custom connectors first.

Cursor and Windsurf

Cursor uses url in .cursor/mcp.json; Windsurf uses serverUrl.

{
  "mcpServers": {
    "philidor": {
      "url": "https://mcp.philidor.io/api/mcp"
    }
  }
}

What the server exposes

search_vaults: Filter vaults by chain, protocol, curator, asset, risk tier, or TVL.
get_vault: Full detail for one vault by network and address.
get_vault_risk_breakdown: Per-vector score breakdown behind a rating.
compare_vaults: Side-by-side comparison of two or three vaults.
find_safest_vaults: Highest-rated vaults for a given asset or chain.
get_protocol_info: Protocol-level coverage and risk posture.
get_curator_info: Curator track record and managed vaults.
get_market_overview: Platform-wide TVL and risk-tier distribution.
explain_risk_score: Plain-language explanation of how a score was derived.
list_vaults_with_incidents: Vaults with recent critical incidents.

Plus three resources (methodology, supported chains, supported protocols) and three guided prompts for due diligence, portfolio review, and yield comparison. Full reference in the MCP documentation.

Questions

Do I need an API key to use the Philidor MCP server?

No. The hosted server at mcp.philidor.io is open and requires no API key or account. Paid Philidor plans exist for direct REST API access, but the MCP tools work anonymously.

Which AI assistants support the Philidor MCP server?

Claude Code, Claude Desktop, ChatGPT (with Developer Mode enabled), Cursor, and Windsurf. Any client that speaks the Model Context Protocol over Streamable HTTP can connect.

What data does it cover?

Independent risk scores and tiers for 800+ DeFi vaults across 10 chains and 11 protocol modules, plus curators, protocols, incidents, and the published scoring methodology.

How are the risk scores calculated?

Every vault is scored 0-10 on a documented four-vector framework covering asset composition, platform and strategy, control and governance, and history. Scores map to three tiers: Prime at 8.0 and above, Core from 5.0 to 7.9, and Edge below 5.0. The full methodology is public.

Need the data directly?

The MCP server is for AI assistants. Teams integrating risk data into their own systems use the REST API.

View pricing API reference