mega-tools

A remote MCP server. Four tools, one home. It is mine, it is yours to use, and it stays where it is.

The tools

ToolWhat it doesInputResult
memo_write Saves a note on the server and returns its id. { text } { id }
memo_search Finds saved notes whose text contains the query. Case-insensitive substring match. An empty query returns every note, newest first. { query } { notes: [{ id, text, createdAt, updatedAt }] }
upload Stores a file on the server and returns a link anyone can open. Give the file as url (the server downloads it), content (text), content_base64 (bytes), or path (a file already in the server data folder). { url | content | content_base64 | path, filename? } { url, id, filename, bytes }
transcribe Turns audio or video at a link into timestamped text. The server downloads the file, transcribes it locally, and returns the segments. start is in seconds from the beginning. { url } { segments: [{ start, end, text }] }

Every tool declares an input schema and an output schema, and every answer comes back as structured content that matches the output schema it declares.

How sign-in works

Only a client I approved can call the tools. The sign-in is OAuth 2.1 with PKCE, run by this server itself:

The metadata for this lives at https://mcp.srv1421711.hstgr.cloud/.well-known/oauth-protected-resource and https://mcp.srv1421711.hstgr.cloud/.well-known/oauth-authorization-server. This page is named there as the resource_documentation of the resource.

What data it keeps

Nothing is shared with a third party. Transcription runs on the machine that serves this page; no audio leaves it.

The protocol

This server speaks the current MCP revision, 2026-07-28: stateless, with the protocol version in the _meta of every request and server/discover to advertise what it supports. It also answers the older, initialize-based revisions so older clients still work. Versions: 2026-07-28, 2025-11-25, 2025-06-18.