Give your autonomous agents ears in the Horn of Africa. The Andebet MCP Server connects your audio files, transcripts, and semantic corpus directly to Claude Desktop and Cursor using the open Model Context Protocol.
Open source NPM package · Compatible with Anthropic Claude & Cursor · Zero setup cost
// ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"andebet": {
"command": "npx",
"args": ["-y", "@andebet/mcp-server"],
"env": {
"ANDEBET_API_KEY": "and_live_7f89b21a99..."
}
}
}
}Autonomous AI agents excel at synthesizing documents, debugging code, and analyzing structured databases. But when you ask Claude or Cursor to analyze an audio file of an Ethiopian parliamentary hearing, a farmer field interview, or a regional radio broadcast, they hit a dead end.
Until now, developers had to write custom glue code: download files, convert codecs, upload to speech services, chunk text, and format prompts manually. It was brittle, slow, and impossible to use directly inside chat interfaces.
The Andebet MCP Server standardizes this completely. By running our lightweight open-source MCP daemon, your agent gains native tools to listen, transcribe, diarize, and query Ethiopian audio files with zero friction.
Standardized agentic connectivity with zero microservice development required.
Paste the 6-line JSON configuration into your Claude Desktop or Cursor MCP settings file with your Andebet API key.
On startup, your LLM client queries Andebet's MCP server and automatically registers speech-to-text, diarization, and corpus search tools.
Ask Claude: 'Transcribe this local Amharic audio recording and write a summary.' The agent calls Andebet and provides verified answers.
Exposed native tools and prompts conform to the official Model Context Protocol schema.
Takes an audio URL or local file path and returns high-accuracy verbatim text with speaker diarization and language codes.
Performs semantic vector search across all audio recordings in your project, returning cited moments and timestamps.
Extracts recognized ministries, figures, monetary figures, and locations in both Ge'ez fidel and Latin spellings.
Isolates distinct speakers across conversations with turn-taking durations and speech percentages.
Exposes project audio collections as readable MCP resources (andebet://projects/{id}/transcripts).
Provides pre-engineered prompt templates for Ethiopian audio summarization, translation, and compliance auditing.
“Plugging Andebet into Claude Desktop via MCP felt like giving Claude a superpower. Now I drag local Amharic recordings into Claude, and it references exact minutes and seconds in its answers.”
Common questions about building agentic workflows with the Andebet MCP Server.
Model Context Protocol (MCP) is an open standard open-sourced by Anthropic that enables AI assistants (such as Claude Desktop and Cursor) to connect directly to local tools, databases, and APIs without requiring custom glue code or proxy servers.
Add a short configuration snippet to your claude_desktop_config.json file specifying 'npx -y @andebet/mcp-server' and passing your ANDEBET_API_KEY as an environment variable. Claude Desktop will automatically detect the server and show the new speech tools upon next launch.
Yes. When running via stdio transport, Claude can point directly to local MP3, WAV, or M4A audio files located in permitted directories on your machine. The MCP server reads the file and streams it securely to Andebet's acoustic engine.
Yes. In Cursor's Features > MCP settings, add the Andebet command. Cursor can then transcribe user test recordings, parse voice bug reports, or search through company audio corpora while you code.
No. The MCP server runs locally as a lightweight stateless bridge between your AI client and our private, SOC-2 compliant processing clusters. All data transmission uses TLS 1.3 encryption.
Yes. In addition to standard stdio mode, you can run the server in HTTP/SSE daemon mode for team environments, centralized proxy instances, and Kubernetes clusters.
Install the open source Andebet MCP Server in less than two minutes and empower your LLM with real speech intelligence.