tensorlake

Tensorlake MCP Server exposes a set of tools that bridge MCP hosts with Tensorlake's document parsing capabilities. After starting the server under the configured path, you can upload documents using upload_document, which accepts a URL, a local file path, or a data URI and returns a document_id. You can then parse the uploaded document with parse_document to convert it into structured data; you can run parsing synchronously or in the background using the sync flag. List_documents lets you view all documents in the current session along with their metadata, and delete_document removes a document by its document_id. The server also exposes a tensorlake://documents resource that provides programmatic access to all documents and their metadata for integration in MCP workflows. This combination enables end-to-end document ingestion, parsing, and management within any MCP host like Claude Desktop, Cursor, or other MCP-enabled clients.

Prerequisites:

• Go programming language installed (go 1.18+)
• Git installed
• Basic familiarity with MCP hosts (Claude Desktop, Cursor, etc.)

Installation steps:

1. Install the binary from the repository:

go install github.com/sixt/tensorlake-mcp@latest

2. Locate the binary (the README suggests it will be named tensorlake-mcp in your GOPATH/bin). For example:

$ where tensorlake-mcp

/Users/<username>/go/bin/tensorlake-mcp

3. Build from source (optional):

# Clone the repository
git clone <repository-url>
cd tensorlake-mcp

# Build the server
go build -o tensorlake-mcp .

The binary will be created as tensorlake-mcp in the current directory. 4) Run the server by providing the required environment variables via the MCP config (see mcp_config below).

Tips and notes:

• You must provide a Tensorlake API key via the TENSORLAKE_API_KEY environment variable. Optionally, set TENSORLAKE_API_BASE_URL if you are using a non-default Tensorlake API endpoint.
• Ensure the MCP host you use supports the MCP server configuration pattern shown in mcp_config. The server will be addressed by the host using the path specified in command.
• When testing locally, you can use the MCP Inspector to validate interactions: npx @modelcontextprotocol/inspector /absolute/path/to/tensorlake-mcp
• If you regenerate the binary or change its location, update the command path in the MCP config accordingly.
• For production deployments, consider securing the API key and restricting network access to the Tensorlake API as appropriate.