Lunara Intelligence

What the Lunara Intelligence MCP server does

The Lunara Intelligence MCP server gives MCP-compatible agents access to Lunara Society’s regulatory corpus. The corpus tracks obligations for the EU AI Act and California SB 942, including the instrument involved, the article establishing an obligation or date, and a link to primary law.

The service is intended for questions where regulatory timing and evidence matter. For example, it can expose the application date for a tracked obligation and the source supporting that date. The published material identifies machine-readable marking under EU AI Act Article 50(2) as applying to generative systems already on the market on 2 December 2026. Applications should treat dates as inputs for their own time-based reasoning rather than assuming the corpus stores a current status.

How it works

Requests are sent as JSON-RPC POST requests to the hosted MCP endpoint:

https://luiqtimzcsoqnizybifs.supabase.co/functions/v1/lunara-mcp

The Lunara Intelligence MCP server exposes four tools: lunara_obligations, lunara_applicability, lunara_cite, and lunara_verify. Their names reflect the server’s four documented functions: retrieving tracked obligations, examining applicability, obtaining citations, and verifying source-backed information.

The server does not bundle a local copy of the corpus. It also does not respond when the authority is unreachable, rather than serving a cached table that might contain an amended date. The source repository identifies mcp/server.mjs as the complete server implementation and provides a real-stdio smoke test for development.

Setup and configuration

Running the hosted service requires an MCP client that can send JSON-RPC requests to the endpoint. No API key, account, or authentication setup is required. The implementation is described as requiring Node.js 18 or later and having zero dependencies. The npm package @lunara/mcp has been written and tested but is not published, so the documented integration path is the hosted endpoint rather than an npm installation.

No environment variables are documented for this server. Clients should configure the endpoint as an HTTP MCP connection according to their own configuration format. The repository does not provide client-specific setup instructions.

Tools and capabilities

The Lunara Intelligence MCP server provides:

• lunara_obligations for accessing tracked regulatory obligations.
• lunara_applicability for consulting the published applicability model.
• lunara_cite for retrieving supporting citations and primary-law references.
• lunara_verify for checking source-backed regulatory information.

The related public files include the obligations corpus, its schema, an applicability model, and an index. These are available independently at lunarasociety.com, which can help developers inspect the data model before connecting an agent.

Limitations and notes

The server’s coverage is limited to the obligations maintained by Lunara Society, specifically the EU AI Act and California SB 942 material described in the source. The README does not claim comprehensive coverage of all AI regulations. It also does not document client-specific behavior, tool argument schemas, or a published npm distribution.

Because no cached corpus is served when the authority is unavailable, requests may fail or return no answer during an outage. Applications that need a durable offline fallback would need to design that behavior separately, while considering the risk that stored dates may become outdated.