Skip to main content

Feature Request: Model Context Protocol (MCP) Support

Feature Request

Implement support for the Model Context Protocol (MCP) within T3 Chat. This would allow T3 Chat to connect to MCP servers, enabling access to a wider range of data sources and tools, and potentially improving the quality and relevance of responses.

Benefits:

  • Enhanced Contextual Awareness: By connecting to MCP servers, T3 Chat could access relevant information from various sources (e.g., Google Drive, Slack, GitHub, Brave Search) to provide more informed and context-aware responses, and become slightly agentic, which brings me to the next point.

  • Improved Agent Capabilities: MCP support would facilitate the development of more sophisticated AI agents within T3 Chat, capable of performing actions and interacting with external systems.

  • Flexibility and Interoperability: MCP's open standard would allow T3 Chat to seamlessly integrate with a growing ecosystem of data sources and tools, and potentially switch between different LLM providers.

  • Data Security: MCP's architecture emphasizes secure data access, ensuring that sensitive information is protected.

Implementation Details:

  • Implement an MCP client within T3 Chat that can connect to MCP servers.

  • Provide a user interface for configuring and managing MCP server connections and credentials/API keys.

  • Develop mechanisms for incorporating data retrieved from MCP servers into T3 Chat's responses and actions.

  • Consider supporting pre-built MCP servers for popular enterprise systems.

  • Also consider having a Desktop Application like Claude where users can run their own MCP servers locally

Use Cases:

  • A user could ask T3 Chat to summarize a document stored in Google Drive, with T3 Chat using MCP to access the document.

  • T3 Chat could be used to create a GitHub issue based on a conversation, using MCP to interact with the GitHub API.

  • T3 Chat could access real-time data from a database via MCP to answer questions about current trends.

Priority:

High - Given the potential benefits for enhancing T3 Chat's capabilities and expanding its integration options.

Additional Notes:

  • Consider leveraging the open-source MCP SDKs (Python, TypeScript, Kotlin) to accelerate development.

  • Explore the example MCP servers and clients provided by Anthropic for inspiration and guidance.

Status: Closed34 comments

Log in to comment and vote

Comments34

  • Yellow Toaster

    This is the #1 feature I’d like to have to be able to replace other chat interfaces (editors/terminal/whatever) for a more agentic workflow. T3 chat would have a huge advantage here because other chat interfaces are one of:
    - Slow
    - Missing syntax highlighting
    - Difficult to read and/or navigate responses in
    And yes, this includes whatever “AI editor” you may be thinking of. Cursor/Windsurf/insert latest VSCode AI fork here, are sluggish. Zed AI chat is just a plain code editor that LLMs can write in and is difficult to read and parse visually.

    All that said… This is probably going to be really tricky to implement since MCP tends to use stdin/stdout as a transport, which obviously does not work in a web app. SSE could probably work, though?

    https://modelcontextprotocol.io/docs/concepts/transports#server-sent-events-sse
    Or really, any custom transport that is not stdio. A WebSocket transport, for example, should be trivial to implement and consume here, but is non-standard, and the entire point of MCP is standards conformance…

    Anyway, if the T3 chat team can figure this out, it would be an absolutely massive productivity boost.

  • Aquamarine Stone

    I can imagine the challenges attaching MCP to a web actor, however this would be an absolutely killer feature. I don’t care about a desktop app unless it’s to implement MCP

    • Coral Deer

      +1 on >I don’t care about a desktop app unless it’s to implement MCP

  • Cyan Spacelab

    Imagine context7 and t3 chat

  • Copper Canary

    If this was added I would instantly swap to T3 Chat.

  • Aquamarine Stone

    SSE implementation would be great, I’ll run my MCP on my own resources.

  • Green Plasma

    MCP is urgently needed thing, it just makes life even more easier. I love MCPs.

  • Emerald Alligator

    Any progress here :)? I use Claude.ai with

    https://github.com/dnakov/claude-mcp and it’s just a pleasure. Would really love to see a MCP server integration here.

    • Emerald Alligator

      Same. I’d be awesome to bring MCP servers to T3Chat - I could see myself using T3Chat a lot if MCP servers were brought here. They’re an essential part of my programming workflow.

  • Purple Kangaroo

    Especially with remote mcp, there is no need for desktop app too.

    https://support.anthropic.com/en/articles/11175166-about-custom-integrations-using-remote-mcp

  • Moccasin Croissant

    It would be appreciated if this comes to T3 Chat, currently searching for dedicated MCP clients to make use of.

  • Apricot Koala

    Just want to add here. I really think this will be big for me especially since I fully utilize Linear via MCP now