Fegis

Define AI tools in YAML with natural language schemas. All tool usage is automatically stored in Qdrant vector database, enabling semantic search, filtering, and memory retrieval across sessions.

python

GitHub Stars

python

Language

4 months ago

First Indexed

3 weeks ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "p-funk-fegis": {
      "command": "uv",
      "args": [
        "--directory",
        "/absolute/path/to/fegis",
        "run",
        "fegis"
      ],
      "env": {
        "AGENT_ID": "claude_desktop",
        "QDRANT_URL": "http://localhost:6333",
        "ARCHETYPE_PATH": "/absolute/path/to/fegis-wip/archetypes/default.yaml",
        "COLLECTION_NAME": "fegis_memory",
        "EMBEDDING_MODEL": "BAAI/bge-small-en"
      }
    }
  }
}

Fegis is an MCP server that lets you define tools in YAML, automatically stores every tool invocation in a semantic memory, and provides powerful search across all tool usage. It helps you build, run, and query structured data workflows with an emphasis on traceability and retrieval.

How to use

You interact with Fegis through an MCP client that can start the local Fe gis server and perform semantic searches across your tool usage. Define tools in YAML, run them, and rely on the built-in memory to recall previous runs and results. Use the SearchMemory-like capability to locate past analyses or outputs quickly, and use the memory context to inform new actions.

How to install

Prerequisites you should have before installing Fegis:

Install the uv package manager and Docker for running Qdrant.

Then follow these steps to set up and run Fe gis locally.

Step by step install and run

# Install uv
# Windows
winget install --id=astral-sh.uv -e

# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# Clone
git clone https://github.com/p-funk/fegis.git

# Start Qdrant
docker run -d --name qdrant -p 6333:6333 -p 6334:6334 qdrant/qdrant:latest

Configure the MCP client for Fe gis

Configure your MCP client to run Fe gis with the following settings. This enables the Fe gis server to access Qdrant, set the embedding model, and identify the agent running in Claude Desktop.

{
  "mcpServers": {
    "fegis": {
      "command": "uv",
      "args": [
        "--directory",
        "/absolute/path/to/fegis",
        "run",
        "fegis"
      ],
      "env": {
        "QDRANT_URL": "http://localhost:6333",
        "QDRANT_API_KEY": "",
        "COLLECTION_NAME": "fegis_memory",
        "EMBEDDING_MODEL": "BAAI/bge-small-en",
        "ARCHETYPE_PATH": "/absolute/path/to/fegis-wip/archetypes/default.yaml",
        "AGENT_ID": "claude_desktop"
      }
    }
  }
}

Notes on usage

Restarting Claude Desktop will expose the Fe gis MCP server under the fe gis name. You will have access to memory-backed search and analysis tools to work with your tool executions.

Configuration

Important environment variables and their purposes are shown in the configuration snippet above. These variables configure the vector store, embedding model, archetype path, and agent identity used by Fe gis.

Security and tips

Keep your Qdrant URL and API key secure. Use a strong, unique API key if you expose Qdrant remotely. Regularly back up your vector data and review access controls for the Fe gis workspace.

Troubleshooting

If Fe gis does not start, verify that Docker is running and that the Qdrant container is healthy. Check that the ARCHETYPE_PATH points to a valid YAML archetype file and that the AGENT_ID is correctly set for your session.

Available tools

BiasDetector

Identify reasoning blind spots, cognitive biases, and systematic errors in AI thinking patterns through structured self-examination

SearchMemory

Store and search tool usage history with semantic similarity and filters

Built by

VeilStrat

AI signals for GTM teams