Get your APIs AI-ready with MCP servers

Since the beginning of Bump, our goal has always been to make APIs usable by humans. We see a new need emerging: APIs now also have to be easily reachable by AI agents.
Thanks to the MCP (Model Context Protocol) standard, there’s now a standardized way to help AI tools interact with applications, APIs, … But building and maintaining an MCP server gets from painful to a burden over time.
By releasing MCP server support, we want to make your MCP server creation as simple as your API docs.
Get a live MCP server running real workflows in minutes #
Drag-and-drop your workflow document on Bump.sh or add a deployment step to your CI, and tada! Your MCP server is ready to run API workflows. Copy the server URL in your favorite AI tool to start using it.
Your MCP servers are based on workflow documents, a deterministic way to describe goals your API users can achieve. These documents can be written using our home-made API workflow standard: Flower. Support for Arazzo (the standard from the OpenAPI Initiative) is coming soon.

Keep control with deterministic MCP servers #
Bump.sh MCP servers expose ready-to-use workflows, unlike most tools that expose individual endpoints. It means that:
- You keep full control of the orchestration: agent requests a result (i.e., book a train trip), and the MCP server follows the workflow you defined.
- AI tools use fewer tokens, thanks to this narrowed-down context.
Private MCP servers? We’ve got you covered #
Like API docs, MCP servers can be private. Manage who gets access directly in Bump. Thanks to our integrated OAuth support, your users authenticate in one click, directly in their AI tools.

Authenticate with your APIs using secrets #
Store sensitive information such as API keys using Secrets, allowing your MCP servers to send requests to authenticated endpoints of your APIs. Secrets stored on Bump are never visible to AI tools, only used to send requests to your APIs.
Debug and optimize your API workflows with debug sessions #
When things go south with API workflows, identifying the root cause can get tricky. We provide you with a dedicated debug endpoint for each one of your MCP servers to help you trace requests and diagnose issues.
Secured on the cloud, or easily hosted in your infrastructure (soon) #
Security is a priority for us. It means that, when using MCP servers hosted on Bump, we don’t store or log any sensitive data sent or received by your MCP servers.
The data plane, responsible for forwarding data between the API and the AI agent, is hosted on a server separated from Bump.sh’s main infrastructure. While using our cloud data plane allows you to quickly get started, we know that data privacy is crucial: that’s why we are working on an option for you to easily host your data plane in your infrastructure, offering you full control over your MCP servers’ data.
Your next step: create your first MCP server #
MCP servers are now available on all our plans. Go to your Bump.sh dashboard to create your first MCP server. Learn more about it in our help center.
As always, don’t hesitate to reach out to us if you have any questions or feedback: we’ll make a lot of improvements over the next months, and it could come from you!