Audiofirst MCP server

MCP (Model Context Protocol) is the standard way to plug tools into AI assistants. Connect any assistant to Audiofirst and it turns any material (notes, word lists, a textbook chapter, formulas) into an Audiofirst pack.

The assistant gets our authoring rules, picks the pack type itself (language cards, formulas or knowledge), asks you only when something is ambiguous, and checks the pack with our validator.

Server URL

https://mcp.audiofirst.org/mcp

No account needed right now. Works with any client that supports remote MCP over Streamable HTTP.

Connect

Claude.ai (web and desktop)

Open Settings → Connectors and choose Add custom connector.
Name: Audiofirst
URL: https://mcp.audiofirst.org/mcp
When asked about sign-in, choose No sign-in, then click Add.
In a chat, open the tools menu and enable the connector if it is not already on.

Claude Code

claude mcp add --transport http audiofirst https://mcp.audiofirst.org/mcp

Other clients (Cursor, VS Code, ChatGPT developer mode and more)

Any client that supports remote MCP over Streamable HTTP. Add this to its MCP configuration:

{
  "mcpServers": {
    "audiofirst": {
      "url": "https://mcp.audiofirst.org/mcp"
    }
  }
}

How to ask

Say the topic, the languages, and roughly how many cards you want. If something is unclear, the assistant will ask. Example prompts:

Make an Audiofirst pack from these notes: …

Build a Swedish→Russian language pack from this word list: …

Turn this physics chapter into a formula pack, with how each formula is read aloud.

Make a knowledge pack of 20 cards from this article: <link or text>

What the server offers

get_authoring_rules: returns our rules for writing a good pack, for each pack type.
validate_pack (read-only): checks a finished pack against our validator and reports problems. It does not save anything.
Rule resources: the same authoring rules, readable by the assistant as MCP resources.

Status

Today the assistant builds and validates the pack. Saving packs straight into your Audiofirst account and publishing to the app and catalog is coming next, together with sign-in for assistants.

Packs are private by default. A pack becomes public only with your consent and after moderation.