The State of API Clients

About this paper

Three previous reports in this series deferred the same company. API management held Postman out as a scope call. API documentation held it out again. API testing held it out a third time, each saying it belonged to its own area. This is that area, and Postman is the reason it looks the way it does.

Across 35 API clients, exactly three publish an API of their own. Postman publishes 35 specifications. Insomnia publishes four. HTTPie publishes three. The other 32 publish nothing, and 30 of the 35 sit at the floor of the agent-readiness scale.

In any other report in this series, a number like that would be a finding about a market failing to keep up. Here it is not. An API client is a consumer of APIs, not a provider of one — and a large, identifiable and growing part of this market is deliberately built with no service surface at all. Yaak advertises “no telemetry and no cloud lock-in.” Bruno and Voiden are “offline-first, Git-native.” Restfox positions itself as “an open-source alternative to Postman.” Four products, four near-identical statements, all defined against the same competitor.

So this report does something the others do not. It scores what can honestly be scored and declines to rank the rest. Six providers operate a hosted multi-user service and therefore have a surface the rubric can read; the membership test is stated plainly so a vendor can check or dispute it. The other 29 are named, tiered and characterized as market evidence — because ranking a command-line HTTP client on commercial clarity would place cURL, among the most widely deployed software ever written, at 20.6, and measure whether a company happens to surround the project.

The scored tier averages 47.6 against a whole-catalog average of 22.8. Postman is the only Exemplar at 75.8 — twenty-two points clear of the next provider, with 35 published specifications and a running MCP server. It has already completed the transition the rest of this market is facing: it stopped being an application and became an API.

That transition is the report’s real subject. An API client is a graphical interface for making HTTP requests, and an agent does not need one. Every other market in this series is being asked to add a machine interface alongside its human one. This one is being asked whether its human interface has a future — and the two halves of the market have given two different, equally coherent answers.

The full report maps the six-provider scored tier facet by facet, the local-first counter-movement and the three forces driving it, the collection-format fight that decides portability in this market, the one place in this series where privacy regulation shapes architecture without mandating anything, and the investable thesis on what happens when the product is a GUI and the customer is a machine.

What's inside

1. Executive summary — the area three reports deferred
2. The state of the market — one platform, and everyone positioned against it
3. Why most of this market cannot be scored
4. The scored tier — six client platforms on the Kin Score
5. Agent readiness — the client that became a platform
6. The local-first counter-movement
7. Provenance — a score with two authors
8. The standards layer — the collection format as the real battleground
9. The regulatory layer — where privacy became a product feature
10. Artifacts for the agentic turn — what happens when the agent replaces the GUI
11. Security and the offline argument
12. Provider-by-provider intelligence
13. The investable thesis
14. Where this is going

What you get for $500.00