◇ HOW IT WORKS

From spec to production MCP server

Point Hintas at your OpenAPI specs. It extracts the multi-step workflows hiding inside, validates them against staging, and deploys a production-ready MCP server your agents call with two tools — search and execute.

247

Endpoints parsed

from a single spec

Workflows extracted

multi-step patterns

MCP tools deployed

search + execute

◇ THE PROBLEM

Exposing raw endpoints to agents doesn't work

One tool per endpoint gives agents verbs with no coordination. They guess ordering, miss failures, and can't roll back.

raw.log

$ curl -X POST /api/accounts/create

→ 201 { "id": "usr_7291" }

$ curl -X POST /api/teams/assign

→ 500 Internal Server Error

$ curl -X POST /api/teams/assign # retry

→ 200 { "team": "engineering" }

$ curl -X POST /api/permissions/grant

→ 403 Forbidden — team not synced

# 4 calls, 2 failures, no rollback strategy

◇ THE TWO TOOLS

Two tools, in detail

Same MCP contract every Hintas deployment ships. Two functions, one safe lane.

Natural-language discovery. The agent describes intent; Hintas returns the matching workflow's id, ordered steps, and required parameters.

hintas.search({
  "query": "Onboard user to engineering team",
  "pattern": "onboard|provision|setup"
})

execute

Validated execution. The agent passes a workflow id and inputs. Hintas runs the sequence, enforces dependencies, and rolls back on failure.

hintas.execute({
  "workflow": "provision-user-onboarding",
  "params": { "userId": "usr_7291", "team": "eng" }
})

◇ TUTORIAL

See it in action

See these workflows in action View pricing plans