← 404.directory

Connect an Agent

One MCP connection for risk preflight before an Agent acts in a prediction market or trusts a third-party tool. No account or API key.

focused toolsetallow / review / blockprivacy-safe identity

Cursor

Open Cursor's MCP installer with a unique non-personal Agent ID already configured.

Add to Cursor →

VS Code / Copilot

Open VS Code's MCP installer with the same privacy-safe attribution setup.

Install in VS Code →

Any stdio MCP client

Use the dependency-free npm bridge when a client or directory accepts only a command. It creates and preserves one privacy-safe Agent ID automatically.

{
  "mcpServers": {
    "404-directory": {
      "command": "npx",
      "args": [
        "-y",
        "@mmvv1638/404-directory-mcp",
        "--source",
        "mcp-registry-showcase.npx"
      ]
    }
  }
}

Claude Code plugin

/plugin marketplace add MM-sheng/404-directory
/plugin install 404-directory@404-directory

Claude Code direct MCP

claude mcp add --transport http --scope user 404-directory https://404.directory/mcp \
  --header "X-404-Agent-ID: agent:a9531d68-0deb-4c70-961b-da1817b667ba" \
  --header "X-404-Source: mcp-registry-showcase.claude-code"

Codex

Add this configuration to ~/.codex/config.toml.

[mcp_servers.404_directory]
url = "https://404.directory/mcp"
http_headers = { "X-404-Agent-ID" = "agent:a9531d68-0deb-4c70-961b-da1817b667ba", "X-404-Source" = "mcp-registry-showcase.codex" }

OpenAI Responses API

Use this request body with your normal OpenAI API authentication and replace the Polymarket URL before sending. The MCP authorization value is a random, non-personal installation token used only for privacy-safe attribution; it does not grant access. The request limits the MCP surface and forces one real evaluate_prediction_market call. OpenAI model usage may incur charges, while 404.directory itself currently requires no account or API key.

{
  "model": "gpt-5.6",
  "input": "Before I act, preflight this real Polymarket market. Market: REPLACE_WITH_EXACT_POLYMARKET_URL. Intended action: observe. Execution mode: supervised. Geographic eligibility: unknown. Return Decision, Reasons, Evidence, Unknowns, and Next action. Do not predict the winner or trade.",
  "tools": [
    {
      "type": "mcp",
      "server_label": "directory_404",
      "server_description": "Preflight Polymarket and third-party Agent actions with evidence-backed allow, review, or block decisions.",
      "server_url": "https://404.directory/mcp",
      "authorization": "agent:a9531d68-0deb-4c70-961b-da1817b667ba@mcp-registry-showcase.openai-responses",
      "allowed_tools": [
        "evaluate_prediction_market"
      ],
      "require_approval": "never"
    }
  ],
  "tool_choice": {
    "type": "mcp",
    "server_label": "directory_404",
    "name": "evaluate_prediction_market"
  }
}

Make the first call useful

Paste an exact market you already care about. Do not use tools/list, a health check, or this homepage as the demo task. If your client exposes MCP Prompts, open preflight-prediction-market.

Copy, replace the URL, and run:
Before I act, use evaluate_prediction_market on REPLACE_WITH_EXACT_POLYMARKET_URL. Set intended_action=observe, execution_mode=supervised, and geographic_eligibility=unknown. Return Decision → Reasons → Evidence → Unknowns → Next action. Do not predict or trade.
Add this Agent policy:
Before a Polymarket observation or contemplated Yes/No action, call evaluate_prediction_market with the exact market, action, approximate notional, execution mode, and the geoblock result from the execution environment. Never treat allow as a prediction. Obey block and pause on review. Report the bounded behavior result with report_prediction_market_outcome.

Installation counts only after a non-error tool result. Connection checks, probes, and repeated calls do not count.

For the direct configurations above, this page generated agent:a9531d68-0deb-4c70-961b-da1817b667ba randomly; keep it stable for that installation. The npm bridge and Claude marketplace plugin create and preserve their own local random IDs. 404.directory stores only an HMAC digest after a successful tool call—never the raw ID, prompt, arguments, or result.