Docs

APIs Are a Mess. So We Open-Sourced a Reference for 3,122 of Them.

MindCloud’s open-source Universal API Reference documents 3,122 apps and 75,075 endpoints in one consistent, searchable format.

For the past five years, I’ve spent a lot of time dealing with the same frustrating reality: integrating with APIs is far harder than it should be.

APIs are supposed to make software work together. In practice, every company does things differently.

One API uses bearer tokens. Another uses OAuth. Another needs a custom header. One paginates with cursors, another with offsets, another with page numbers, and some barely document it at all. Dates, filtering, sorting, errors, rate limits, argument names, response formats—every API makes its own choices.

Even some of the largest companies in the world have APIs that are surprisingly difficult to work with.

That inconsistency creates a bad experience for developers, integration teams, and now AI-agent builders who need software to reliably work across many systems.

So we built a consistent API surface internally to solve the problem for ourselves. Today, we’re making a large part of that work available to everyone.

Introducing the Universal API Reference

We’ve open-sourced the Universal API Reference: a searchable, machine-readable reference covering:

  • 3,122 apps

  • 75,075 endpoints

  • An average of roughly 24 documented endpoints per app

  • One consistent documentation format across every app

  • One file per endpoint

The repository documents the real differences developers run into when working with APIs, including:

  • Authentication requirements

  • Request and response formats

  • Pagination behavior

  • Filtering and sorting conventions

  • HTTP methods and status codes

  • Argument conventions

  • Native vendor endpoint mappings

  • API quirks and links back to source documentation

It is designed to be useful whether you are integrating directly with a vendor’s native API, researching capabilities across many platforms, building developer tools, or giving an AI agent reliable API context.

Explore the Universal API Reference on GitHub →

The API ecosystem has a translation problem

There is no single REST API standard in the real world.

There are conventions, specifications, and best practices. But once you start integrating with real systems, you quickly discover that every vendor has its own interpretation of all of them.

The same basic task—say, finding a customer, creating an invoice, issuing a refund, or fetching records—can look completely different from one API to the next.

That means developers repeatedly have to answer the same questions:

  • How does this API authenticate?

  • How does it paginate results?

  • Can I filter or sort this endpoint?

  • What does the vendor call this object or action?

  • What shape does the response take?

  • What is the native endpoint behind this operation?

  • What exceptions or quirks should I expect?

The Universal API Reference makes those answers easier to find and compare.

Instead of opening dozens of vendor documentation sites and trying to interpret their different conventions, developers can search a consistent body of documentation.

For example:

# Find every app with a refund-related endpoint
rg -l "refund" apis

Or inspect how a specific API handles pagination:

rg "pagination" apis/s/stripe/universal-api/README.md -A 3

That may sound simple, but it changes the experience of researching integrations. It turns API discovery from a scavenger hunt into something you can search, inspect, compare, and build on.

Built from production integration work

This is not scraped documentation assembled from the web.

The Universal API Reference is a nightly mirror of the machine-readable specification that powers MindCloud’s integration gateway. It is the result of years spent building, validating, and maintaining integrations across thousands of applications.

Our production systems execute API calls at a run rate of approximately 1.85 billion API calls per year.

That matters because documentation is only useful when it reflects reality.

When a spec is wrong, an integration breaks. When an API changes, workflows fail. Our team has to identify the issue, understand the vendor’s behavior, and update the underlying specification. Those improvements flow back into this open-source reference through regular updates.

The result is a living resource rather than a static directory.

Useful for developers—even without MindCloud

We want to be clear about what this project is and what it is not.

The Universal API Reference is useful on its own. You do not need a MindCloud account to browse it, search it, use it as a native API reference, or bring the parts you need into your own development and AI workflows.

You can use it to:

Research native APIs faster

Each app includes documentation for its native API operations, including endpoint mappings, arguments, behavior, and links to the vendor’s own documentation.

Search across thousands of APIs

Find APIs that support a capability—refunds, invoices, contacts, subscriptions, webhooks, pagination, custom objects, and more—without starting from scratch on every vendor documentation site.

Build better AI agents

AI agents need structured, current context to interact with software reliably. The repository includes a machine-readable app catalog and an llms.txt index so teams can use the reference in retrieval-augmented generation workflows, agent tools, and internal knowledge systems.

Compare API conventions

Understand how APIs differ before you invest engineering time in an integration. Authentication, pagination, filters, response shapes, and naming conventions are all documented in a consistent structure.

The documentation content is published under a CC BY 4.0 license.

A shared foundation for the next generation of software

The rise of AI agents makes this problem even more important.

An agent that needs to operate across business software cannot succeed on vague descriptions of what an API might do. It needs structured information about what actions exist, what inputs they require, how data is returned, and how the vendor’s system actually behaves.

Developers need that same clarity.

We believe better infrastructure for integrations should not be locked away or unnecessarily difficult to explore. We built this reference because we needed it to create better integrations ourselves. We are sharing it because we think other developers, teams, and builders can benefit from it too.

This is our contribution to making the API ecosystem more navigable.

The hosted Universal API and MCP

The open-source reference is the map.

For teams that want a unified way to execute against these apps—not just research them—MindCloud also offers a hosted Universal API and MCP.

It provides a consistent interface for calling supported applications through one API surface, rather than requiring a separate implementation for each vendor.

But the repository stands on its own. We hope it becomes useful to developers regardless of whether they ever use MindCloud’s hosted platform.

Try the Universal API →

Help us improve it

The Universal API Reference will continue to evolve as APIs change and new apps are added.

If you find an error, see missing coverage, or have a cross-API question you think should be easier to answer, please let us know.

And if you are building something with it, we would love to see it.