MCP Server

Model Context Protocol

Connect any AI assistant (Claude, GPT-4, Cursor…) directly to Mindstk financial data. 53 tools — 17 data tools and 36 analysis-context tools — all direct DB reads, one API key, no quota split.

What is MCP?

The Model Context Protocol is an open standard by Anthropic that lets AI assistants call external tools in a structured, secure way. Instead of pasting data into a conversation, the AI calls a Mindstk tool directly, gets structured financial data back, and uses it to answer your question in real time.

Real financial data

Live prices, fundamentals, ratios, news — straight from the Mindstk database.

Analysis context on demand

Thesis, risk, earnings & sector tools return structured DB context for your model to reason over — no server-side LLM.

One key, no quota split

Every tool is a direct database read. None call an LLM internally, so there is no standard/AI quota distinction.

Tools & access

Every tool is a direct database read — none run an LLM server-side, so there is no quota distinction. They are grouped only by purpose.

Data tools (17)

Market data, fundamentals, ratios, screener, news, watchlists, peer comparisons.

Analysis-context tools (36)

Narrative, thesis, risk, earnings, sector and portfolio tools — return structured DB context for your assistant to reason over (no server-side LLM).

How to connect

In your chat — Claude, ChatGPT, Gemini

No terminal

The easiest way. Add Mindstk as a connector and sign in with your Mindstk account — you never copy an API key. Usage is billed against your plan automatically.

Open your assistant’s settings → Connectors → Add custom connector.
Paste the server URL below.
A Sign in to Mindstk window opens → log in → Authorize.
Done — the Mindstk tools appear in your chat.

Connector URL:

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

Sign-in uses secure OAuth 2.1 — no key to manage, revocable anytime from your account.

For developers — CLI & config files

1. Get your API key

Create an account and generate a key from your API Keys dashboard.

Recommended

Claude Code — .mcp.json

Create a .mcp.json file at the root of your project (or ~/.claude.json for global access). The "type": "http" field is required — without it, auth headers are ignored.

{
  "mcpServers": {
    "mindstk": {
      "type": "http",
      "url": " https://api.mindstk.io/v1/mcp/",
      "headers": {
        "Authorization": "Api-Key YOUR_API_KEY"
      }
    }
  }
}

Or via CLI:

claude mcp add --transport http mindstk  https://api.mindstk.io/v1/mcp/ \
  --header "Authorization: Api-Key YOUR_API_KEY"

Claude Desktop — claude_desktop_config.json

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

{
  "mcpServers": {
    "mindstk": {
      "url": " https://api.mindstk.io/v1/mcp/",
      "headers": {
        "Authorization": "Api-Key YOUR_API_KEY"
      }
    }
  }
}

Server endpoint:

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

Test with curl

List all tools:

curl -X POST  https://api.mindstk.io/v1/mcp/ \
  -H "Authorization: Api-Key YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

Call a tool:

curl -X POST  https://api.mindstk.io/v1/mcp/ \
  -H "Authorization: Api-Key YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 2,
    "method": "tools/call",
    "params": {
      "name": "get_ascore",
      "arguments": { "tickers": "AAPL,NVDA,MSFT" }
    }
  }'

How it works

1
You ask a question
"What is the investment case for NVIDIA given its current valuation?"
2
The AI selects tools
Claude calls get_ascore, get_ratios, and get_investment_rational for NVDA automatically.
3
Mindstk returns real data
The MCP server validates your key and returns live, structured data straight from the database.
4
The AI synthesises
Claude combines structured financial data with its own analysis to give a complete answer.

Tool catalog

(53 tools — click to expand)

Data tools

17 tools

Analysis-context tools

36 tools

Protocol reference

Supported methods

initialize — handshake, server info
tools/list — full tool catalog
tools/call — execute a tool

Error codes

-32700 — Parse error
-32600 — Invalid request
-32601 — Method not found
-32001 — Auth / invalid key
-32005 — Rate limit exceeded
-32603 — Tool execution error

// Tool response format
{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "content": [{ "type": "text", "text": "..." }]
  }
}