Speakeasy
Generate production MCP servers and SDKs from your API, plus an MCP gateway
Platform to discover, install, and run MCP servers and tools
mcp.run is a Dylibso platform for discovering, installing, and running MCP servers as portable, sandboxed Wasm servlets, featuring a dynamic reprogrammable MCP server that lets clients install tools on demand.
mcp.run is part of the fast-growing MCP ecosystem, providing a registry and runtime for MCP servers ('servlets') that give AI assistants access to tools, data, and integrations. Built by Dylibso — known for its WebAssembly tooling — mcp.run packages tools as portable, sandboxed Wasm servlets that can run across environments, addressing security and portability concerns that come with connecting agents to external capabilities. A notable feature is its dynamic, reprogrammable MCP server: rather than statically wiring up every tool, a client can connect once and then discover and install additional MCP servers from mcp.run on demand. This makes it easier to extend an agent's toolset without reconfiguring clients, and the Wasm sandbox model helps contain what installed tools can do. The broader MCP ecosystem grew rapidly through 2026, with the specification reaching a major release and hundreds of servers cataloged across categories. mcp.run positions itself as infrastructure for teams building on MCP who want a curated, portable, and secure way to distribute and run tools for their agents.
mcp.run is a Dylibso platform for discovering, installing, and running MCP servers as portable, sandboxed Wasm servlets, with a dynamic reprogrammable MCP server for on-demand tool installation.
mcp.run is built by Dylibso, a company specializing in WebAssembly developer tooling. It applies that expertise to the MCP ecosystem, distributing tools as portable, sandboxed Wasm servlets.
As MCP adoption accelerated through 2026, mcp.run positioned itself as infrastructure for securely distributing and running agent tools across MCP-compatible clients.
mcp.run provides a registry and runtime for MCP servlets, plus a dynamic, reprogrammable MCP server that lets clients discover and install additional servers on demand. Tools run as sandboxed WebAssembly for portability and safety.
It integrates with MCP clients such as Claude and Cursor, targeting developers who want a curated, secure way to extend agents with external tools.
mcp.run serves developers and teams building on the Model Context Protocol who need a portable, secure way to distribute, install, and run tools for their AI agents.
Developers extending AI agents with MCP tools.
Engineering leads adopting MCP infrastructure for their teams.
MCP ecosystem contributors and WebAssembly practitioners.
Developer teams building on MCP who value portable, sandboxed tool distribution across clients like Claude and Cursor.
mcp.run is a Dylibso product; specific funding details are not prominently public. Verify with the vendor or public sources.
mcp.run is a platform by Dylibso for discovering, installing, and running MCP (Model Context Protocol) servers as portable, sandboxed WebAssembly servlets.
It is a reprogrammable MCP server that lets a connected client discover and install other MCP servers from mcp.run on demand, without reconfiguring the client.
Packaging tools as WebAssembly servlets makes them portable across environments and sandboxed, improving security when agents run third-party tools.
It works with MCP-compatible clients such as Claude and Cursor, and other tools that support the Model Context Protocol.
There is free access to discover and run servlets, with paid team and enterprise options for advanced needs. Verify current pricing with the vendor.
Side-by-side pages for pricing, features, and best-fit use cases.
Generate production MCP servers and SDKs from your API, plus an MCP gateway
Managed integrations and MCP gateway that gives AI agents authenticated tool access.
The Docker Hub of MCP: discover, install, and host thousands of MCP servers.
MCP server registry, in-browser inspector, and gateway with an LLM API bundled in.