Published on 2026-03-02
]]>Published on 2026-03-02
Tags: commentary, llm, protocol
I’m going to make a bold claim: MCP is already dying. We may not fully realize it yet, but the signs are there. OpenClaw doesn’t support it. Pi doesn’t support it. And for good reason.
I agree. I tried the Github MCP once, watched as my naive granting of privileges resulted in massive context usage (each permission becoming an exposed API), and never went back. As Armin Ronacher said, CLI tools and regular code suffice. Like Eric, I think MCP slowly fades and most of the companies who built MCP servers deprecate them.
]]>Published on 2025-11-20
]]>Published on 2025-11-20
Tags: javascript, protocol, til
Today at work I learned of the existence of Server Sent Events (SSE) and some transport layer/protocol facts about Websockets.
SSE is just a standardization of the old Comet model. The browser initiates an HTTP request, the server and the client keep the underlying TCP connection alive, with the server sending events-as-they-happen by adding more data onto a neverending HTTP response body.
Websockets also start life as an HTTP request to the server, but the initial request is to switch protocols, after which the server and client drop down to communicating in a binary frame-based protocol on top of the underlying TCP connection.
]]>Published on 2025-11-16
]]>Published on 2025-11-16
Tags: commentary, llm, protocol, software-eng
]]>The fact that MCP is a difference surface from your normal API allows you to ship MUCH faster to MCP. This has been unlocked by inference at runtime
Normal APIs are promises to developers, because developer commit code that relies on those APIs, and then walk away. If you break the API, you break the promise, and you break that code. This means a developer gets woken up at 2am to fix the code
But MCP servers are called my LLMs which dynamically read the spec every time, which allow us to constantly change the MCP server. It doesn't matter! We haven't made any promises. The LLM can figure it out afresh every time
Published on 2025-09-04
]]>Published on 2025-09-04
Tags: commentary, llm, protocol, software, software-eng, text-editor
You can now run Claude Code directly in Zed and use it side-by-side with Zed's first-party agent, Gemini CLI, and any other ACP-compatible agent.
As a follow up to the Agent Client Protocol post, apparently Zed went with both prongs of the strategy — helping get ACP into another text editor, AND integrating ACP into another coding agent.
]]>Published on 2025-08-27
]]>Published on 2025-08-27
Tags: commentary, llm, protocol, software, software-eng, text-editor
]]>You can now interact with third-party agents directly within Zed. To make this possible, we created the Agent Client Protocol (ACP), and we've partnered with Google to integrate Gemini CLI as the initial reference implementation.
...
Command-line agents are cool because their simplicity makes them easy to run anywhere—including as a subprocess of another application. Zed was already running Gemini CLI inside our embedded terminal emulator, but we needed a more structured way of communicating than ANSI escape codes. So we defined a minimal set of JSON-RPC endpoints to relay user requests to the agent and render its responses. The result is the Agent Client Protocol, a lean framework that lets any client talk to any agent, as long as they follow the schema.
Published on 2025-08-18
]]>Published on 2025-08-18
Tags: commentary, llm, ocaml, plt, protocol, software-eng
A general challenge with MCP today is that the more tools you have, the more you’re contributing to context rot. You’re also limited to rather low amounts of input. On the other hand, if you have an MCP that exposes a programming language, it also indirectly exposes a lot of functionality that it knows from its training.
I've been thinking about how one could design a programming language, knowing that a language model would be used in the crafting of programs in said programming language. Models have trouble generating nested "language", e.g. outputting code inside of JSON (and failing to make the code syntactically correct), or as Armin documents their poor ability at using MCP (data and semantics embedded in JSON).
]]>