README.md

metadata

title: Aileen3 Core
emoji: 👩🏻‍💼
colorFrom: purple
colorTo: blue
sdk: docker
pinned: false
license: cc-by-4.0
short_description: Aileen 3 Core - Information Foraging MCP
tags:
  - building-mcp-track-enterprise
  - building-mcp-track-customer

Aileen 3 Core

Introduction

Large Language Models (LLMs) rely on tools - sometimes provided by MCP servers - to interact with the outside world. Aileen 3 Core is an MCP server that focuses on Information Foraging: mining for novel insights from high-noise sources to create dense briefings for time efficient consumption by the user. Grounded in cognitive science, Aileen 3 models novelty as prediction error against explicit priors such as user expectations, facts from an AI Memory Bank, or media context. To that end, the Aileen 3 Core MCP server provides media access and analysis services backed by Google Gemini.

Competition submission

Aileen 3 Core is a contender in the MCP's 1st Birthday - Hosted by Anthropic and Gradio hackathon. Aileen 3 Agent, an agentic system built on this MCP server, is a capstone project to the AI Agents Intensive Course with Google.

Using

Using in Claude Desktop

Installing

1. Optionally, create a new Python virtual environment

• example: python3 -m venv .venv-claude

2. Install the Aileen MCP: pip install ./mcp

• (or pip install -e ./mcp if you want to make live changes to this source tree)

3. Obtain Google Gemini API key: [https://aistudio.google.com](Google AI Studio)
4. Add reference to claude_desktop_config.json. The Gemini API key will be read from the environment, so can be set here:

{
...
  "mcpServers": {
    "aileen3-mcp": {
      "command": "/Users/.../aileen3-core/.venv-claude/bin/python",
      "args": [
        "-m",
        "aileen3_mcp.server"
      ],
      "env": {
        "GEMINI_API_KEY": "AI..."
      }
    }
  }
}

4. Restart Claude

Using the MCP server

The model Haiku 4.5 is sufficient for basic tasks. To make your plans fully transparent to the LLM, refer to "aileen3" explicitely in the prompt, e.g.:

Use aileen3 to translate slide 3 from YouTube video reference eXP-PvKcI9A to German.

Debugging

The message exchange and Claude-facing error messages can be read from Claude log files:

tail -n 20 -F ~/Library/Logs/Claude/mcp*.log

Local development

Build and run the Docker Space image locally:

docker build -t aileen3-core .
docker run -it -p 7860:7860 aileen3-core

Troubleshooting

• Error message: "google.genai.errors.ClientError: 401 UNAUTHENTICATED. {'error': {'code': 401, 'message': 'API keys are not supported by this API. Expected OAuth2 access token or other authentication credentials that assert a principal. ...". This may be a catch-all error message and the underlying problem may be something different. We solved this by using an older Gemini API key from Gemini Studio (Nov 10 vs. Nov 23; same project). The newer API keys may have been generated by Vertex AI, but we did not investigate this further. The working key starts with "AI", the broken keys start with "AQ.".