Skylit

What Skylit MCP server does

The Skylit MCP server exposes read-only access to Skylit's data intelligence platform for US markets. Its coverage includes options flow, volatility, dealer positioning, gamma and vanna heatmaps, and dark pool activity. The server is hosted by Skylit; this repository contains client setup guides, examples, SDK references, and synchronization checks rather than the server implementation itself.

The catalog contains 68 tools. They support discovery, flow analysis, screeners, options chains, market-wide measures, sector flow, historical comparisons, volatility analysis, dark pool queries, and account usage. A separate intelligence toolset combines several underlying data calls into concise responses for common agent questions.

How it works

Clients connect over Streamable HTTP to https://mcp.skylit.ai/mcp. Authentication can use Skylit's OAuth sign-in flow or an API key sent as a bearer token. Once connected, an MCP client can invoke read-only tools and receive results containing fields such as a factual summary, data, and metadata. Tool metadata can include credit cost, remaining balance, timestamp information, and whether a result is partial.

The standard endpoint exposes the full catalog. Agents that need only higher-level answers can use the intelligence endpoint with ?toolset=intelligence. That list includes explain_levels, vol_context, whats_changed, market_brief, and flow_context. These tools address questions such as key dealer-positioning levels, implied-volatility context, changes since a specified time, broad market conditions, and a ticker's current options-flow read.

The Skylit MCP server uses account credits for calls. One credit equals $0.001, typical calls cost between one and five credits, and failed 4xx or 5xx requests are refunded. The account_usage tool does not consume credits. Intelligence tools have their own listed costs, so agents should account for usage when choosing between detailed calls and combined responses.

Setup and configuration

Add the hosted URL to an MCP client that supports Streamable HTTP. Clients may use different configuration field names, so the appropriate client-specific setup guide should be followed. With JSON-based configuration, the server URL can be paired with an Authorization: Bearer YOUR_SKYLIT_API_KEY header. Omitting the header allows OAuth sign-in with Skylit where the client supports it.

Access is currently limited to Skylit members with API access, and the project identifies the API and MCP service as beta. API keys are obtained from Skylit's Developer page. The Python SDK reads SKYLIT_API_KEY, while the TypeScript SDK uses the same credential convention; these SDKs are separate ways to access the platform and are not required for an MCP client connection.

Tools and capabilities

The available groups include:

• Flow search, feeds, aggregates, scores, sweeps, momentum, strikes, and tide analysis
• Daily and weekly screeners for underlyings and contracts, plus unusual volume and open interest
• Bull/bear ratios, chain and contract statistics, historical comparisons, moneyness, and volatility/open-interest measures
• Option chains, charts, relative-volume tools, market breadth, market tide, and sector flow
• Dark pool trades and top prints
• Heatseeker gamma and vanna heatmaps, levels, historical heatmaps, and related statistics
• Tempest volatility tools covering implied volatility, term structure, cones, surfaces, events, snapshots, screeners, history, and derived measures

Limitations and notes

OHLCV price bars are not available through MCP; the README identifies Atlas REST endpoints for that data. Tempest tools depend on the user's plan and return not_entitled without access, without charging for those responses. The repository's catalog and REST specifications are checked against the live API periodically, so tool availability may change as the beta service evolves.