The thing that stands out is that everything comes from one place. Connect a repo and the docs, Postman collection, mock server and MCP server all get generated from the same API contract — so you don't end up with three descriptions of the same endpoint disagreeing with each other. Setup took a few minutes and the output was usable without much cleanup. The MCP server generation is the part I hadn't seen elsewhere: it turns an API catalog into something an agent can actually browse and call, not just something a person reads.
Hey Product Hunt, Ana here 👋
Four years ago, I was fed up with Postman and wanted to replace it. I started with docs as a side project. That became Theneo. We got into YC, and today 20,000+ companies use it.
Through Theneo, we’ve seen how messy API management gets. Now some of our customers get more API calls from agents than humans, while the way they manage those APIs has barely changed.
That’s why we built Elva. We started with a few customers, and today we’re opening it up to everyone.
You can connect your repo and Elva builds your API catalog, maps how it’s used, and checks AI readiness (fyi no spec required).
From there, create contracts around actual use cases: choose the endpoints and fields a customer, partner, or agent needs. Publish to docs platforms like Theneo or Mintlify, mocks, or hosted MCP servers. When your code changes, review the updates or sync them automatically under your publishing rules.
For MCP, we handle hosting and auth. You can test across models, see which tools are slow or failing, and collect feedback directly from agents when they get stuck.
Would love for you to try it with your own repo. Does it cover how you actually manage your APIs? What are we missing?
@rahulkumaran4 Thank you! That’s exactly how we think about it, AX is becoming as important as DX. Elva evaluates factors such as tool discoverability, schema clarity, parameter descriptions, authentication flows, response consistency, error handling, context efficiency, and whether an agent can reliably complete real tasks end to end.
It’s less about simply validating an MCP server and more about testing whether an agent can understand the available tools, choose the right one, recover from failure, and reach the intended outcome. We’re continuing to expand both the criteria and task-based evaluations as agent patterns evolve.
Solid MCP host. When one server serves different audiences, how do you keep auth and usage analytics separate?
@pratiksharda Thanks! We treat each audience as a distinct tenant or client boundary, even when they share the same MCP server. Each gets its own authentication policy, scopes, and credentials, with authorization enforced per tool and resource, not just at the server level.
For analytics evry request carries a tenant/client identifier through the full execution path. That lets us track usage, latency, errors, and quotas separately without exposing sensitive user data. We also keep audit logs scoped by tenant and apply strict redaction and retention policies. In short: shared infrastructure, isolated identity, permissions, quotas, and observability.
@pratiksharda Thanks for your question. We’ve designed Elva so shared MCP infrastructure doesn’t mean shared trust or visibility. The separation is enforced at the consumer level, with clear attribution and controls throughout. I look forward to hearing your feedback once you get a chance to throughly explore Elva!
Do you support agent keys for different agents, so you can see which agent did what?
@naimz hey yes we do track which agent and user is using tool and how, which tools are failing, and even if something fails you get a feedback from the agent directly
@naimz thanks for the interesting question. I would love to hear your feedback once you get a chance to throughly explore Elva :)!
It is pretty straightforward to set up and get running. I like the MCP setup and the way it lets you test how the API works with agents. Seems like a useful tool for teams working on making their APIs agent-ready. Congrats on the launch!
@sophie_gurjishvili Thanks Sophie, let us know how it holds up in production
@sophie_gurjishvili thanks so much! I would love to hear more when you get a chance to thoroughly explore it
Solid product. Is there a CLI for it? If so, how can I use it. Can't wait to integrate this into my current workflow.
@mark_smith_1 yes, we have a CLI, its in our docs + we have more updates coming from our CLI + incorporating with your agents
@mark_smith_1 Indeed! You can find a detailed guide in our docs. Just let me know if you are unable to find what you are looking for, I would be happy to give you a detailed demo.
Congrats Ana and team! 🎉
As someone who builds full-stack projects, I know how quickly API stuff can get messy as a project grows. I like that Elva lets you connect the repo and get things organized from the code itself, instead of maintaining everything separately.
Definitely going to try it on one of my projects!