Hi I'm Jonah and I'm the CEO of Layer! Our team has been building in the space of connecting LLMs to external APIs and tools for almost a year now and we're psyched about the recent traction around MCP. MCP is an exciting protocol, but there is still a lot of tooling that needs to be built around it to turn it from a "cool toy" into a legitimately useful protocol. Our platform is an early attempt at addressing this. To elaborate a bit, a big part of the challenge when it comes to MCP is reliability. If you expose your entire API through MCP, LLM clients like Cursor and Windsurf get overwhelmed and fail to use your API reliably. Our platform attempts to solve some of these problems around MCP and make it a more reliable and scalable standard for giving context/tools to LLMs. Here is the flow of how it works: 1. You upload your OpenAPI spec on Layer. 2. Our platform "cleans" your spec until it is "MCP ready". This involves removing unwanted routes, handling authentication, and more. 3. You specify which endpoints to expose via MCP and how these endpoints can be orchestrated together. We call this feature workflows, and it give you a deeper level of control over how your MCP server behaves against complex queries and user demands. 4. We give you ongoing insights and telemetry data on how users are actually using your MCP server. We're excited to see what you think of the platform and will be around to answer any questions!

Layer

Build an MCP server in minutes

Build an MCP server in minutes