llm-mcp-client 0.1a0
Simon Willison's Weblog Simon Willison ● Covered by 7 sources
Simon Willison dropped an early alpha of llm-mcp-client, a plugin adding MCP support to his llm command-line tool. It means llm users can soon hook into any Model Context Protocol server instead of waiting on custom integrations.
Based on reporting by Simon Willison's Weblog, Simon Willison — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Simon Willison quietly pushed out version 0.1a0 of llm-mcp-client on July 31st, and if you've been watching where the LLM tooling world is headed, this one's worth a second look. The plugin bolts Model Context Protocol support onto his widely used llm command-line tool, the same one that already lets you chat with dozens of model providers from a terminal. MCP, if you haven't run into it yet, is the open standard Anthropic put out for letting language models talk to external tools and data sources in a consistent way, rather than every app inventing its own glue code.
What makes this interesting is less the code itself, which is explicitly marked as an alpha release, and more what it signals about llm's trajectory. Willison built llm as a Swiss-army knife for interacting with models locally, and MCP support turns that knife into something closer to a universal remote. Instead of writing bespoke plugins for every tool a model might need to call, you point llm at an MCP server and the tool-calling machinery just works. That's the pitch, anyway.
Early alpha means early alpha, though. A 0.1a0 version number is Willison being honest that this is a first pass, likely missing edge cases and probably going to change shape before it settles. Anyone who has used his other llm plugins knows he iterates fast and in public, often shipping something rough and then refining it based on how people actually use it. Expect this one to follow the same pattern over the coming weeks.
The bigger context here is that MCP itself is still young, having only emerged in late 2024, and adoption has been uneven across the ecosystem. A client implementation from someone with Willison's reach — his llm tool has become something of a default for developers experimenting with models outside a browser — gives MCP another foothold outside of the big lab-backed integrations. Whether it becomes the way people connect llm to external tools, or just one option among several, will depend on how quickly the alpha rough edges get sanded down.
My take — AI-written commentary, not fact-checked reporting
I like that Willison ships these things half-finished and lets the community kick the tires, because that's how the useful stuff in this ecosystem actually gets built — not through some vendor's polished roadmap. MCP needs exactly this kind of scrappy, independent tooling to avoid becoming another standard that only the model labs bother to implement well.
Read more about this at: Simon Willison's Weblog
