What is the Model Context Protocol?

MCP is an open standard from Anthropic that lets AI assistants connect to external data sources and tools through a uniform interface. Servers advertise their capabilities once; compliant clients call them without custom integration.

Which MCP clients does the OAS server work with?

ChatGPT, Claude, Perplexity, and Grok. The server follows the MCP spec; client-specific setup is in the GitHub README.

Do I need a separate API key for the MCP server?

No. The MCP server is included with Pro and API subscriptions and authenticates with your Options Analysis Suite email and password - OAuth for Claude, ChatGPT, and Grok, and base64-encoded credentials for Perplexity. No separate API key is required. The Python SDK and REST API are a separate programmatic surface that does use an API key on the $99/mo API tier.

Is the MCP server open source?

The canonical version lives in the OAS monorepo and is deployed alongside the API. A public mirror on GitHub tracks the same source so users can audit the tool definitions, inspect the wire protocol, and propose changes.

← Developers Hub

MCP Server

Options Analytics as Native AI Tools

The Model Context Protocol (MCP) server gives ChatGPT, Claude, Perplexity, and Grok direct access to 44 options analytics tools: pricing models, Greeks, dealer flow, screeners, market regime, and the full market and regulatory data layer. Included with a Pro subscription. No API key required; sign in with your Options Analysis Suite credentials. See the AI Assistant feature documentation for the full tool inventory and example questions.

View on GitHub → · AI Assistant Docs · Start 14-Day Trial

What is MCP?

Model Context Protocol is an open standard published by Anthropic in late 2024 that lets AI assistants connect to external data sources and tools through a uniform interface. Instead of every integration being custom (plugin APIs, function-calling schemas, retrieval pipelines), MCP defines one wire protocol for tool discovery, tool execution, and context exchange. Servers like the OAS MCP advertise their capabilities once; supported clients can call them.

Practically: instead of an LLM guessing options-pricing answers from training data, it queries the OAS engine in real time and gets the same authoritative numbers the web app shows.

44 tools across three categories

The platform's market, research and regulatory data layer, your synced analyses, and Black-Scholes pricing are exposed through MCP tools the AI can call. The assistant reads that data layer and reasons over it in seconds, instead of you clicking through every chart. The other pricing models, calibration and multi-model runs stay on the REST API and Python SDK.

Market, research and pricing (37 tools). Pricing-model history across the 17 engines (Black-Scholes, Heston, SABR, Variance Gamma, Jump Diffusion, Local Volatility, Monte Carlo, FFT, PDE, Binomial CRR, plus 7 exotic engines: Asian, Barrier, Lookback, Digital, Compound, Chooser, Multi-Asset), full 17-Greek output, IV surfaces and history, options chains, stock prices, a unified regime tool with market/per-symbol/intraday scopes (net gamma/delta/vega/vanna/charm/vomma exposures, walls, gamma flip) plus the VIX term structure, sector and industry P/E by exchange and CFTC futures positioning, a unified options screener covering 16 screens (most-active, unusual, highest IV rank, exposure leaders, VRP, max pain, term-structure backwardation, put skew, model divergence, regime stress, pre-earnings IV, day-over-day leaderboards, call/put unusual splits), market-wide trends, earnings calendars, company profiles with ESG scores and fund holdings, fundamentals with peer P/E and market-cap history, analyst data, news, insider filings (per company and market-wide), institutional ownership from 13F filings and the ETFs holding a stock, Congress trades and activist filings, SEC EDGAR filings with merger and offering flags, dark pool data, a unified FINRA short-side tool (daily volume + biweekly interest), fail-to-deliver, threshold-list history, dividends, stock splits, a unified market calendar (economic, IPO, dividend, split), trading halts, and a unified Treasury rates tool (benchmark + curve). Also end-of-day dealer positioning for roughly 5,500 tickers and the end-of-day options snapshot for roughly 6,000 tickers, and, on Pro and above, a live options chain, live dealer positioning and live strategy scans read from the broker connected to your account, the eight-model calibration fit history, and Black-Scholes pricing from explicit inputs.
Your synced data (6 tools). Pricing-model run history, FFT scanner results, AI Compute Suite runs, GEX snapshots, portfolio snapshots, risk snapshots, analysis rollups, and filtered analysis queries. Available when Data Sync is enabled in Account then AI Settings.
Platform context (1 tool). Background knowledge about pricing models, Greek definitions, and data sources so the AI interprets your results accurately rather than reasoning from generic training data.

See the AI assistant feature page for the full per-tool inventory and example questions.

Setup: Claude (remote MCP connector)

No download required; works in any browser at claude.ai.

Open Claude at claude.ai (Claude Pro subscription required).
Click your profile icon in the bottom-left, then go to Settings then Integrations.
Under MCP Connectors, click Add custom connector.
Set Name to Options Analysis Suite.
Set URL to https://mcp.optionsanalysissuite.com/mcp.
Leave OAuth Client ID and OAuth Client Secret empty.
Click Add. When prompted, sign in with your Options Analysis Suite email and password to authorize.

Setup: ChatGPT (one-click app)

Install the Options Analysis Suite app from the ChatGPT app directory; works in ChatGPT's web app and desktop app.

Open the Options Analysis Suite app in ChatGPT: chatgpt.com/apps/options-analysis-suite.
Sign in with your Options Analysis Suite email and password to authorize. The 44 tools are now available in any chat, in both the web app and desktop app.

Setup: Perplexity (Streamable HTTP)

Perplexity uses an API-key authentication shape; encode your credentials as base64.

Open Perplexity (desktop app or web browser).
Go to Settings then MCP Connectors then Add Custom Connector.
Set Name to Options Analysis Suite.
Set MCP server URL to https://mcp.optionsanalysissuite.com/mcp.
Set Authentication to API Key.
In a terminal, encode your credentials: echo -n "your-email:your-password" | base64. Paste the resulting string as the API key.
Set Transport to Streamable HTTP.
Click Add. Enable the connector from your Connectors list, then select it as a source when starting a new thread.

Setup: Grok (Custom Connector)

OAuth flow; same pattern as Claude.

Open Grok at grok.com.
Go to Connectors then New Connector then Custom Connector.
Set Name to Options Analysis Suite.
Set the MCP server URL to https://mcp.optionsanalysissuite.com/mcp.
Save the connector. When prompted, sign in with your Options Analysis Suite email and password to authorize.
Optional: enable Data Sync in AI Settings for access to the 6 synced-data tools.

Enabling personal data sync (optional)

Six of the 44 tools read from your synced account data; the other 38 work without sync. Pricing analyses, FFT scans, AI Compute Suite runs, GEX snapshots, portfolio snapshots, risk snapshots, analysis rollups, and the strategy definitions you send from the Strategy page are all sync-gated. Toggle Enable Data Sync on the AI Settings page and future analyses, scans, and snapshots become available to the AI automatically. Sync is opt-in and can be disabled at any time. Without sync, the 37 market, research and pricing tools and the platform-context tool still work normally.

Why MCP matters for options analytics

Options pricing is a domain where LLMs routinely guess wrong from training data. Asking ChatGPT for "the gamma exposure on SPY" without an MCP tool produces an authoritative-sounding number with no actual source, and the answer is usually stale, sometimes by months. With MCP, the same question routes through OAS and returns the real number the webapp's GEX page renders. The AI gets to be useful for live options analysis instead of being confidently wrong.

The deeper value: AI assistants can now reason over OAS data the way a quant would. Show me the tickers where dealer gamma is concentrated below current spot. Find names with steep put skew and IV rank above 70. Compare 30-day calibrated Heston vol with realized vol across my watchlist. Each one is minutes of manual chart-clicking, answered in seconds by an assistant that can read the full analytics surface.

Authentication and pricing

The MCP server is included with a Pro (Professional) subscription, which includes a 14-day free trial. The MCP server is not gated to the separate API tier; any Pro subscriber can connect Claude, ChatGPT, Perplexity, or Grok. Authentication is your Options Analysis Suite email and password: OAuth for Claude (Desktop and Web), ChatGPT, and Grok; a base64-encoded email:password string for Perplexity's API-key field. No separate API key is issued. End-of-day market data is included; real-time data requires a free BYOK broker integration (Tradier, tastytrade, Public, or Schwab).

Open development

The MCP server has two repositories: the canonical version lives in the OAS monorepo (private) and is deployed alongside the API; a public mirror on GitHub at Options-Analysis-Suite/options-analysis-suite-mcp tracks the same source for transparency and so users can inspect the tool definitions, audit the wire protocol, and propose changes via issues and pull requests.