Trading Volatility MCP

The options & volatility data source agents can query.

A Model Context Protocol server for Trading Volatility — lets any AI agent discover and retrieve options market-structure data (GEX, gamma flip levels, dealer positioning, skew, max pain, expected-move levels, options flow, and ranked trade setups) over the public v2 API, in conversation.

• Read-only — discovers and retrieves data over your existing Trading Volatility subscription. No orders, payments, or monitoring.

• Stateless passthrough — forwards your Authorization: Bearer <key> to the v2 API and stores nothing. Local and hosted modes are the same code path.

• Two transports — stdio for local agents, streamable HTTP + SSE for remote, multi-user hosting.

• Works without a key — the demo tickers (AAPL, VIX, KO, META, AMZN, XOM, GM, MCD) work out of the box.

Tools

Related MCP server: E*TRADE MCP Server

Quickstart (local, stdio)

# Run straight from PyPI with uvx (or: pipx run tv-mcp)
uvx tv-mcp                       # stdio; uses TV_API_KEY or a config file

# …or from source
python3 -m venv .venv && . .venv/bin/activate
pip install -e ".[dev]"
cp config.example.json config.json   # add your API key (git-ignored)
python -m tv_mcp                      # stdio by default

Validate:

ruff check . && pytest

Credentials & precedence

The key is resolved in this order (first match wins):

1. Per-request header — Authorization: Bearer <key> (or X-Api-Key) — hosted mode.

2. Environment — TV_API_KEY.

3. Local JSON config — tv_api_key in config.json (see config.example.json; values support ${ENV_VAR} substitution so the key can stay in the environment).

4. Demo mode — no key; only the demo tickers are available.

The key is never logged or persisted.

Remote (hosted, HTTP + SSE)

TV_MCP_TRANSPORT=http PORT=8000 python -m tv_mcp   # serves http://0.0.0.0:8000/mcp

Each request carries its own key, so one deployment serves many users:

POST /mcp           Authorization: Bearer <your-key>
GET  /health        liveness probe
GET  /AGENTS.md      agent-discovery doc (how to use this server)

Container build (binds $PORT, runs the HTTP transport — deploys to any container host such as Cloud Run, Fly, or ECS):

docker build -t tv-mcp .
docker run -p 8080:8080 tv-mcp

The server is stateless and holds no secrets, so it scales horizontally with no extra setup; tune limits with the env vars in .env.example.

Connecting an agent

Claude Code / Claude Desktop (local, stdio) — claude_desktop_config.json:

{
  "mcpServers": {
    "trading-volatility": {
      "command": "uvx",
      "args": ["tv-mcp"],
      "env": { "TV_API_KEY": "your-key" }
    }
  }
}

Remote MCP clients (Claude web custom integrations, OpenAI Responses mcp tool, Gemini function-calling, the Vercel AI SDK, ADK's MCPToolset, LangChain/LangGraph's MultiServerMCPClient) all point at the same endpoint and pass the key as a header:

• URL: https://<your-deployment>/mcp

• Header: Authorization: Bearer <your-key>

Because the server is a standard streamable-HTTP MCP endpoint with header auth, no per-client shim is needed — configure the URL and header in whichever framework you use. Copy-pasteable guides with real code for each:

• Anthropic Claude — Claude Code, Desktop, the Messages API MCP connector, and Claude web

• OpenAI · Google Gemini · Google ADK · LangChain/LangGraph · Vercel AI SDK

• Index: docs/integrations/

How it works

agent ──tools──▶  TV MCP  ──HTTPS (Bearer key)──▶  stocks.tradingvolatility.net/api/v2
                 (stateless)

The agent carries continuity between turns; the server keeps no session state. It forwards the caller's key and returns the v2 payloads unchanged (they are already agent-shaped).

Repository layout

src/tv_mcp/
  cli.py          stdio | http entry point
  server.py       FastMCP wiring (tools, resources, /health, /AGENTS.md)
  settings.py     config loading + precedence
  auth.py         credential resolution (header → env → config → demo)
  tv/             v2 API client + normalized errors
  tools/          one module per tool group (tickers, curves, discovery, auth)
  transports/     stateless HTTP + SSE app, per-request key middleware
tests/            client, auth, settings, tools, transport, smoke
docs/             design, build plan, implementation notes

Docs

• docs/integrations/ — per-framework integration guides (Claude, OpenAI, Gemini, ADK, LangChain, AI SDK)

Deployment runbooks and maintainer planning artifacts are kept internal and excluded from public releases.

License

MIT — see LICENSE. Open source under the Trading Volatility brand.