Introducing the Query Agent MCP Server
Read the Portuguese translation →
Author: Weaviate vector database · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: Introducing the Query Agent MCP Server · Channel: Weaviate vector database · 03:56 · Watch on YouTube
Topics: AI agents, AI developer tools
Summary
Section titled “Summary”Weaviate introduces a native MCP server for its Query Agent, explaining how tool-calling agents connect to external services to gather context and sometimes make changes. By hosting and updating the server, Weaviate removes the need for users to build, deploy, scale, and maintain their own tool server. The Query Agent MCP can connect with tools and frameworks including Claude Code, Pydantic AI, LangChain, and DSPy. The speaker notes that the server’s tool descriptions and inputs may change, and recommends checking the documentation for current details.
Key points
Section titled “Key points”- The Query Agent now supports a native MCP server.
- Tool-calling agents connect language models to external services to gather context.
- Some agent systems can also make write requests to external services.
- Building and maintaining tool servers can require describing, deploying, scaling, and updating them.
- Weaviate hosts and updates the Query Agent MCP server.
- The server connects to MCP clients and can be used with tools and frameworks such as Claude Code, Pydantic AI, LangChain, and DSPy.
- The server describes ask mode and search mode tools, their input arguments, and their response format.
- The speaker advises checking the documentation for up-to-date implementation details.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:08 | Introducing the Query Agent MCP server | The speaker introduces the native MCP server and explains how tool-calling agents use external services for context. |
| 00:38 | Why host the MCP server natively | Weaviate’s hosted server removes the need for users to build, deploy, scale, and maintain their own tool servers. |
| 01:10 | Connecting to AI tools and frameworks | The speaker gives examples of connecting the Query Agent MCP to Claude Code, Pydantic AI, LangChain, and DSPy. |
| 01:42 | Tool details and documentation | The video introduces the server’s ask and search tools and notes that viewers should consult the documentation for current details. |
Quotes
Section titled “Quotes”Weaviate’s Query Agent now supports a native MCP server. — 00:08
The first generation of two building tool calling agents often required you to build, deploy, scale, and then probably most annoyingly maintain the tool servers yourself. — 00:38
We host it, we update it, and you don’t need to worry about this overhead with accessing the Weaviate Query Agent. — 01:10
please check out the docs for a more accurate for the particular details of how we’re doing this. — 02:14
Concepts and entities
Section titled “Concepts and entities”Weaviate, Query Agent, MCP server, tool calling loop, Claude Code, Pydantic AI, LangChain, DSPy
Full transcript (original language)
Section titled “Full transcript (original language)”00:08 Hey everyone, I’m super excited to share that Weaviate’s Query Agent now supports a native MCP server. This video will start with why we built it and how to use it, and then dive into the details. Thank you so much for watching. I really hope you find this useful. The tool calling loop has become one of the most common abstractions for building AI agents. This is where a large language model is connected to one or multiple external services that it sends requests to to gather context. The latest systems even allow write access as well. So, it might send read requests to GitHub, Slack, the Query Agent, and Notion, and
00:38 then maybe send a write request to Slack. Let’s start with the value of having the MCP server natively built into the service. The first generation of two building tool calling agents often required you to build, deploy, scale, and then probably most annoyingly maintain the tool servers yourself. So, again, to connect your agent to external services, you need to figure out how you want to describe the tool, how you’re going to deploy it, scale it, and then update it with new versions of the software you want to connect your agents
01:10 to. Compared to this, the Query Agent MCP just directly connects to your MCP clients. We host it, we update it, and you don’t need to worry about this overhead with accessing the Weaviate Query Agent. Here are some quick examples showing how easy it is to connect the Query Agent MCP to common AI tools like Claude Code or agent frameworks like Pydantic AI, LangChain, or DSPy.
01:42 Claude Code Pydantic AI LangChain India’s Pie Now, let’s dive further into the details of how the Query Agent MCP server will describe the ask mode and search mode tools, the input arguments, and response
02:14 format to your agent. Now, before diving into this, this is recorded around October 2026, and one of the key values of us maintaining the server is that we’ll be updating this description and inputs and so on. So, for if you’re watching this in the future, please check out the docs for a more accurate for the particular details of how we’re doing this.
03:56 >> Thank you so much for watching this video explaining the Query Agent MCP server. I really hope that you find this interesting and can imagine ways that this can help with the software that you’re building. If so, please email us at agentsteam@wevia.io. We’d love to learn more about your use case and see how we can work together and how we can help. Thanks again.