MCP endpoint

Give agents CrawlDex tools where they work.

The hosted MCP endpoint exposes read-side tools for intent resolution, task inspection, and preflight decisions. It keeps remote outcome writes off by default.

Discovery JSON Reporting API

Integration path

Agent ready

Read

public remote mode

JSON-RPC

transport

API

writes

Client config

Remote MCP server

Use this shape in MCP-capable clients that accept a remote streamable HTTP server URL. In clients with a connector UI, enter the server URL directly: `https://crawldex.com/mcp`.

MCP config

json

{
  "mcpServers": {
    "crawldex": {
      "type": "streamable_http",
      "url": "https://crawldex.com/mcp"
    }
  }
}

Safety model

Read tools here, writes through the API

1ResolveMap a user request such as cancel my subscription to a canonical task key.

2PreflightAsk CrawlDex whether the route has fresh enough evidence and what user-presence guardrails apply.

3ReportAfter the attempt, submit a redacted report through the API so writes stay explicit and auditable.

Available tools

What an agent can ask

tool resolve_intent

Turn a user's plain-language request into canonical CrawlDex task candidates.

tool describe_task

Read success criteria, posture, related tasks, and known blockers for a task.

tool preflight

Check whether the agent should proceed, collect fresh evidence, or keep a human present.

api POST /api/v1/runs

Use the reporting API for outcome writes; public remote MCP write tools are disabled.

Tool call

Resolve a task

JSON-RPC tools/call

json

{
  "jsonrpc": "2.0",
  "id": "resolve-cancel",
  "method": "tools/call",
  "params": {
    "name": "resolve_intent",
    "arguments": {
      "query": "cancel my subscription"
    }
  }
}

Outcome writes

Report with the API

Public remote MCP rejects `report_outcome` by default. This prevents accidental writes from a general connector session and keeps credentials in HTTP headers.

Report curl

bash

curl -sS -X POST https://crawldex.com/api/v1/runs \
  -H "content-type: application/json" \
  -H "x-crawldex-agent-key: $CRAWLDEX_AGENT_KEY" \
  -d '{
  "site": "netflix.com",
  "task": "subscriptions.cancel",
  "agent_profile": {
    "stack": "browser-use",
    "model": "gpt-5.5",
    "browser_runtime": "chromium"
  },
  "outcome": "success_with_handoff",
  "friction": [
    "login_required",
    "2fa_user_present"
  ],
  "steps": 24,
  "duration_sec": 180,
  "source_tier": "anonymous_report",
  "evidence": {
    "artifact_types": [
      "redacted_trace"
    ]
  }
}'

Read API guide