Postman introduced Orbit, a discovery layer that turns a task description into the right API calls, and the framing is blunt about why it exists: “An agent that can reach the right API can send the invoice, pull the customer record, or page the on-call engineer. An agent that can’t is a chat window.” It ships two tools in sequence. search takes a query up to 512 characters and returns up to 25 candidate public endpoints with a fitness grade; integrate takes up to 10 of them and returns a task brief with the auth, the request steps, the response codes and the parameter dependencies. The headline number is the context cost — roughly 2,500 tokens for a brief against roughly 69,000 for PayPal’s OpenAPI, a 27x reduction — at a median six seconds for search and eleven for integrate. It is live at POST /v1/search and POST /v1/integrate, as a REST API or an MCP server, with no key and no signup.
The 27x is measured against one large specification, so read it as an illustration rather than a constant — the ratio for a small API would be unremarkable. Three design decisions survive independent of the number. Splitting search from integrate keeps endpoint selection in front of a human instead of buried inside a tool call. The FIT grade is a stated verdict rather than a ranking — “‘Fully’ means go. ‘Partially’ means you’re one endpoint short.” And the third is a commitment most vendors avoid putting in writing: “Orbit doesn’t pad the gap with a plausible-looking email call it doesn’t have, and it doesn’t quietly drop the half of your request it can’t cover.” A discovery tool promising to return less rather than invent more is the correct promise, and it is a testable one.
Postman carries 51 API pages in the catalog, and Orbit is a sibling to two of them rather than an extension of either: the Postman MCP Server is the existing agent-facing door onto the workspace, while the Postman apiCatalog API is where an organisation’s own inventory lives. Orbit points outward at public APIs instead, publishing a fitness judgment about how well an arbitrary API serves an agent task — the same question the Agent Readiness dimensions ask, arrived at from the integration side rather than the contract side.
Postman scores 67.1, exemplar on the Kin Score — among the strongest in the catalog — on discoverability at 83.3, access and commercial clarity at 76.3, developer ergonomics at 73.8, with contract governance at 28.8 the clear laggard. Agent Readiness is 39.1, agent-ready, and the lit row is ten dimensions deep, including error_semantics, event_surface_described, agent_skills and well_known_catalog. Orbit needing no auth at all is the right call for a read-only discovery tool, and it also means the harder half is still ahead. delegated_identity, dynamic_client_registration and protected_resource_metadata are unlit on Postman’s own surface, alongside idempotency and reversibility_documented. Finding the right call is the part Orbit just made cheap. Letting an agent make it on someone else’s behalf, safely and reversibly, is the part nobody has made cheap yet.