# weaviate/mcp-server-weaviate [Health: Active]

**Category:** 🧠 Knowledge & Memory  
**Repository:** https://github.com/weaviate/mcp-server-weaviate  
**GitHub Stars:** 161  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/weaviate-mcp-server-weaviate

## Description
An MCP Server to connect to your Weaviate collections as a knowledge base as well as using Weaviate as a chat memory store.

## Tools
Capabilities this server exposes over MCP:

- **weaviate-collections-get-config** — Inspect collection schemas
- **weaviate-tenants-list** — List tenants in a multi-tenant collection
- **weaviate-query-hybrid** — Hybrid (vector + keyword) search
- **weaviate-objects-upsert** — Create or update objects

## Claude Desktop Quick Installation
No confirmed install details yet, so no config is published. Follow the setup steps in the repository above.

## Documentation

## What weaviate/mcp-server-weaviate does

The weaviate/mcp-server-weaviate MCP server connects MCP clients to Weaviate collections and exposes a focused set of collection, tenant, search, and object-management operations. Agents can inspect collection schemas, retrieve tenants for multi-tenant collections, run hybrid searches that combine vector and keyword retrieval, and create or update objects.

The repository is no longer the recommended deployment path. Its earlier standalone Go implementation wrapped the Weaviate REST API, but that implementation has been superseded by an MCP server included in the main Weaviate distribution.

## How it works

The current implementation runs inside the Weaviate `weaviate/weaviate` binary rather than as a separate process from this repository. Weaviate exposes the MCP endpoint at `/v1/mcp`, using the same port as its REST API. Authentication uses Weaviate's existing API-key flow, and the endpoint observes Weaviate RBAC controls.

The built-in server is available as a preview beginning with Weaviate v1.37.1. MCP clients connect to the endpoint, and the server performs the corresponding operations against the configured Weaviate instance and collections.

## Setup and configuration

Use the built-in Weaviate MCP server instead of installing or running this standalone repository. Enable it by setting the following environment variable for the Weaviate process:

```sh
MCP_SERVER_ENABLED=true
```

After enabling the feature, connect to the `/v1/mcp` endpoint on the Weaviate REST API port. API-key authentication and RBAC are inherited from Weaviate, so access depends on the credentials and permissions configured for that instance. The repository excerpt points to Weaviate's MCP documentation for detailed setup, environment-variable, RBAC, and per-tool reference information.

## Tools and capabilities

The weaviate/mcp-server-weaviate MCP server provides four documented tools:

- `weaviate-collections-get-config` inspects collection schemas and configuration.
- `weaviate-tenants-list` lists tenants belonging to a multi-tenant collection.
- `weaviate-query-hybrid` performs hybrid retrieval using vector and keyword search.
- `weaviate-objects-upsert` creates new objects or updates existing objects.

Together, these tools support basic agent access to a Weaviate-backed knowledge base and object writes. The documented tool list does not describe broader administration, deletion, or collection-creation operations.

## Limitations and notes

The standalone repository is deprecated and is no longer maintained. There is nothing to install or run separately from this repository for the current implementation. Use a Weaviate release that includes the preview MCP server and enable it in the main binary instead.

The endpoint also depends on Weaviate's existing authentication and RBAC behavior, so an MCP connection does not bypass those controls. The repository history remains available for reference, but earlier standalone source should not be treated as the current implementation.

_Full upstream README: https://allmcps.com/mcp/weaviate-mcp-server-weaviate/readme_