mcp -weaviate

This MCP server integrates with the Weaviate vector database to expose common MCP tools for manipulating and querying objects. The repository provides two core tools: Insert One, which inserts a single object into Weaviate, and Query, which retrieves objects using Weaviate's hybrid search capabilities. The server expects you to run the containerized MCP server and connect your MCP clients to it, enabling you to exercise insert and query operations through the MCP protocol. The included test client target helps validate basic operations by sending requests that correspond to the Insert One and Query endpoints.

Prerequisites:

• Docker installed and running
• Access to the Weaviate instance (if not bundled with the MCP server)

Installation steps:

1. Pull and run the MCP server image (as defined in the mcp_config):

   docker pull weaviate-mcp-server-weaviate docker run -i weaviate-mcp-server-weaviate

2. Ensure network connectivity between the MCP server container and your Weaviate instance. If Weaviate is on localhost, expose the necessary ports or use a docker network bridge.

3. If you are using a test client, ensure your environment can reach the MCP server endpoint (the container’s exposed port or a mapped port on the host).

4. (Optional) Build from source if you prefer not to use the prebuilt image. The repository README references build steps (make build) and a test client (make run-client). You can run these locally if you clone the repo and have the necessary toolchains installed.

Typical commands referenced in the repository:

• Build the server: make build
• Run the test client: make run-client

Notes and tips:

• The server provides two tools: Insert One and Query. Use Insert One to add a single object to Weaviate, and use Query to perform hybrid searches across text and vector representations.
• If you encounter connectivity issues, verify that the MCP server container can reach Weaviate (check hostnames, IPs, and ports) and that any required API keys or authentication are configured.
• When testing locally, you may need to configure environment variables for the Weaviate endpoint, authentication, or index name. Placeholder environment variables can be added in mcp_config under env if needed by your deployment.
• If you update the server image or environment, rebuild and re-run the container to pick up changes.
• If you run into tool-specific errors, consult the Weaviate documentation for hybrid search parameters and payload formats, since the Insert One and Query operations will rely on the appropriate schema existing in Weaviate.