Getting started with CORE3 MCP

This guide walks you through connecting an MCP client to core3, the public CORE3 MCP server. Once connected, your agent can call read-only tools for category quality scores, project metrics, Proof of Voice, and exchange risk data.

Server URL

https://api.core3.io/v1/mcp

Transport

CORE3 MCP uses Streamable HTTP—a stateless HTTP transport defined by the MCP specification. Your client opens a session against the endpoint URL; no WebSocket or stdio process is required.

Authentication

No authentication is required. You do not need an API key or x-api-key header to use MCP tools.

If you need the full REST API surface, search, ratings pagination, or commercial rate limits, request an API key via CORE3 for partners or contact [email protected].

Conventions

slug — Canonical project or exchange identifier passed to most tools. Brand names often differ from the slug (for example Compound → compound-governance-token, Coinbase → coinbase-exchange). Prefer resolve_project_slug when the user gives a brand, ticker, or name. Use get_projects_list when you need the full project catalog. There is no exchange discovery tool—only call exchange score tools with a confirmed CORE3 slug.
days — Positive integer for history tools (get_project_categories_scores_dynamic, get_exchange_categories_scores_dynamic). Specifies how many days of score history to return.

Connect your client

Cursor

Add the server to your MCP configuration file (project-level .cursor/mcp.json or user-level settings):

cursor-mcp.json

{
  "mcpServers": {
    "core3": {
      "url": "https://api.core3.io/v1/mcp"
    }
  }
}

Restart Cursor or reload MCP servers after saving the file. The server appears as core3 in your agent’s available tools.

Claude Desktop

Add the server to claude_desktop_config.json:

claude_desktop_config.json

{
  "mcpServers": {
    "core3": {
      "url": "https://api.core3.io/v1/mcp"
    }
  }
}

Restart Claude Desktop for the configuration to take effect.

Other MCP clients

Any client that supports MCP Streamable HTTP can connect using the same URL. Refer to your client’s documentation for the exact configuration format.

Your first tool calls

After connecting, verify the integration with these tools in order:

1. Check health

Call check_health with no parameters. A successful response confirms the server is reachable:

The service is healthy

2. Resolve a project slug

Call resolve_project_slug with query: "ethereum". On success, structuredContent.status is resolved and you get resolvedSlug / resolvedName for follow-up tools. If status is ambiguous or not_found, do not guess a slug or answer for a different entity—ask the user to choose or provide the exact CORE3 slug. Optionally call get_projects_list to browse the full catalog (structuredContent.projects and totalProjects).

3. Fetch a category quality score

Call get_project_security_score with slug: "ethereum". The response includes:

content — Human-readable confirmation text for the agent
structuredContent — Machine-readable JSON with envelope fields and the score object

Example structured payload shape:

{
  "status": "ok",
  "entityType": "project",
  "metric": "security_score",
  "requestedSlug": "ethereum",
  "resolvedSlug": "ethereum",
  "scoreType": "category_quality",
  "scoreDirection": "higher_is_better",
  "interpretationHint": "This is a category quality/calculation score, not overall PoL. Higher scores are better (stronger quality). Lower scores are worse. Never treat a high category score as higher risk.",
  "score": {
    "score": 86.6,
    "rating": "AAA",
    "confidence": "High"
  }
}

rating is a letter grade (AAA–D). confidence is Exceptional, High, Moderate, or Low. See Tools reference for every tool’s return shape.

Response format

Every tool (except check_health) returns two layers:

Field	Purpose
`content`	Text summary the agent reads during conversation
`structuredContent`	JSON payload your integration or agent logic can parse

Agents typically surface the text summary to users while using structuredContent for downstream reasoning or display. Detail and score tools include a shared envelope (status, entityType, metric, slug fields, and—for scores—scoreType / scoreDirection).

Troubleshooting

Issue	What to try
Server not appearing in client	Confirm the URL is exactly `https://api.core3.io/v1/mcp` and restart the client
Tool returns an error for a slug	Call `resolve_project_slug` first; do not continue after `ambiguous` or `not_found`
CORS errors in browser-based clients	Use a supported MCP client rather than calling the endpoint directly from a browser
Need data not exposed as a tool	Use the Projects Data API or Exchanges Data API—search, ratings, overall PoL, and other endpoints require REST

Need full REST API access, higher rate limits, or commercial terms? Contact [email protected]

Next steps

Browse the full Tools reference for all 23 tools, parameters, and REST equivalents.
Read MCP overview for scope and limitations.
Pair category score results with Project PoL methodology or CEX PoL methodology for interpretation.

Getting Started

API Reference

MCP

Probability of Loss (PoL)

Project PoL

CEX PoL

Trust & Transparency

CORE3 for You

About CORE3

Getting started with CORE3 MCP

Server URL

Transport

Authentication

Conventions

Connect your client

Cursor

Claude Desktop

Other MCP clients

Your first tool calls

1. Check health

2. Resolve a project slug

3. Fetch a category quality score

Response format

Troubleshooting

Next steps

​Server URL

​Transport

​Authentication

​Conventions

​Connect your client

​Cursor

​Claude Desktop

​Other MCP clients

​Your first tool calls

​1. Check health

​2. Resolve a project slug

​3. Fetch a category quality score

​Response format

​Troubleshooting

​Next steps

Server URL

Transport

Authentication

Conventions

Connect your client

Cursor

Claude Desktop

Other MCP clients

Your first tool calls

1. Check health

2. Resolve a project slug

3. Fetch a category quality score

Response format

Troubleshooting

Next steps