Model Context Protocol 2.0 reignites agent infrastructure momentum

The specific draw here is the stateless design of MCP 2.0, which is a meaningful departure from earlier session-based approaches. Stateless protocols are easier to deploy, cache, and scale, which is why Willison found it worth building new tooling around rather than treating this as an incremental patch.

Modelwire has no prior coverage of MCP or agent tooling standards in the archive, so this sits largely disconnected from anything we've tracked before. That absence is itself worth noting: the agent infrastructure layer, meaning the plumbing that lets LLMs call external tools reliably, has been developing quickly with relatively little mainstream editorial attention. Willison's decision to build mcp-explorer and datasette-mcp in direct response to the spec update is a useful signal that the protocol has crossed a threshold where third-party developers find it stable enough to build on, not just experiment with.

Watch whether other prominent open-source developers ship MCP 2.0-compatible tooling within the next 60 days. Broad third-party adoption at that pace would confirm the stateless redesign solved the friction that slowed earlier uptake, while silence would suggest the barrier is still implementation complexity rather than protocol design.