Weaviate MCP Server – Vector Data and Docs

Weaviate provides official MCP support for both live database access and documentation retrieval. Use the built-in Weaviate MCP server to inspect schemas, search data, and modify objects in a Weaviate instance, or use the hosted Weaviate Docs MCP server to retrieve current Weaviate documentation inside an AI coding environment.

#vector-database#hybrid-search#documentation

Overview

Weaviate provides official MCP support for two related workflows. The built-in
Weaviate MCP server lets AI assistants interact directly with a Weaviate
database instance. The hosted Weaviate Docs MCP server gives AI coding tools
live access to Weaviate documentation, tutorials, API references, GitHub issues,
and community content through a Kapa.ai-powered knowledge engine.

What the MCP server enables

The built-in Weaviate MCP server exposes live database operations. Depending on
the instance configuration and client permissions, an AI agent can:

Inspect Weaviate schemas.
Search data with vector and hybrid search.
Retrieve relevant objects from collections.
Modify objects in the connected Weaviate instance.
Use Weaviate as a live retrieval source for agentic or RAG workflows.

The Weaviate Docs MCP server is separate and focused on documentation. It helps
agents find current guidance for schema design, hybrid search, client-library
usage, deployment, modules, and other Weaviate development tasks.

When to use it

Use the built-in Weaviate MCP server when an agent needs direct access to a
real Weaviate instance, such as debugging retrieval quality, checking a
collection schema, searching application data, or updating objects during a
supervised workflow. Use the docs server when a coding agent needs current
Weaviate examples or API guidance and should avoid hallucinated or outdated
client-library patterns.

Connection and authentication

The built-in server runs on the same port as the Weaviate REST API at
/v1/mcp; for a default local instance, the endpoint is
http://localhost:8080/v1/mcp. It is disabled by default and must be enabled
with MCP_SERVER_ENABLED=true.

The hosted docs server is available at https://weaviate-docs.mcp.kapa.ai.
Both servers use Streamable HTTP and work with MCP-capable clients such as
Claude Code, Claude Desktop, Cursor, VS Code, and ChatGPT Desktop. Authentication
for the built-in database server follows the Weaviate instance's own security
configuration.

Key considerations

The built-in Weaviate MCP server can access and modify real database objects, so
connect it only to trusted clients and enable it deliberately. Use development
or staging data for experimentation, enforce the same authentication and
network controls used for the underlying Weaviate instance, and require review
before object-modifying actions. The docs server is suitable for AI-assisted
coding, but it is documentation retrieval only and does not operate on a user's
database. Public docs-server access may be rate-limited, so retry later if
queries temporarily return no results.

Frequently Asked Questions

When should an AI agent use the Weaviate MCP server?

Use the built-in server when an agent needs live access to a Weaviate instance for schema inspection, vector or hybrid search, object retrieval, or supervised object modification. Use the docs server when the agent needs current Weaviate documentation for code generation or troubleshooting.

What does Weaviate MCP add to an AI agent's capabilities?

It gives the agent either live access to a Weaviate database through the built-in MCP endpoint or up-to-date Weaviate documentation through the hosted docs server, reducing reliance on stale model knowledge and manually copied console or documentation content.

What can an AI agent access or manage through Weaviate MCP?

Through the built-in database server, the agent can inspect schemas, search data with vector or hybrid retrieval, and modify objects in the connected Weaviate instance. Through the docs server, it can search Weaviate documentation, tutorials, API references, GitHub issues, and community content.

How is authentication configured for Weaviate MCP?

The built-in server uses the authentication and authorization configuration of the Weaviate instance it runs inside. The docs server is a hosted documentation endpoint. Do not hardcode database credentials in MCP client files, and protect the built-in endpoint with the same controls used for the Weaviate REST API.

Which transport should be used for Weaviate MCP?

Use Streamable HTTP. Connect to /v1/mcp on the target Weaviate instance for live database operations, or connect to https://weaviate-docs.mcp.kapa.ai when the agent only needs Weaviate documentation.