Skip to main content

How AI agents discover and transact with governance services


For an AI agent to use a governance service, two things have to be true: the service has to be discoverable — the agent can find out what it does and how to call it without a human reading docs — and it has to be transactable — the agent can complete the work, including payment, programmatically. OntoRamp is built for both.

Everything below is served from machine-readable surfaces on ontoramp.com — no scraping, no guessing.


Discovery: the machine-readable surfaces

An agent starts by reading the descriptors. Each one is a stable URL that enumerates capabilities, endpoints, and tiers:

llms.txtPlain-text capability summary: what OntoRamp does, the tool catalogue, service tiers, and engagement URLs.
.well-known/mcp.jsonThe MCP front-door descriptor — every tool an agent can call, with endpoints and auth.
agents.jsonThe purchasable plugin catalogue, drift-guarded to the on-site pricing.
openapi.yamlOpenAPI 3.1 for the MCP tool calls and the checkout REST endpoint.
.well-known/ai-plugin.jsonAI-plugin manifest for clients that consume the legacy plugin format.
sitemap.xmlEvery public route, for traditional crawlers and generative-AI indexers.

Calling the tools

The MCP front-door exposes 15 tools across 4 plugins — knowledge-graph search, governance-maturity evaluation, projection, and decision intelligence.

After discovery, an agent registers a free key on the MCP plugin hub and calls the governance tools over the MCP Streamable HTTP endpoint. Authentication is a Bearer key — see agent authentication for the exact challenge — and every parameter and response shape is in the MCP API reference.


Transacting: programmatic checkout

An agent does not need a human to buy a plugin. POST a plan lookup_key to the checkout endpoint — no API key required — and OntoRamp returns a Stripe Checkout link to complete payment. The operation is documented as createCheckoutLink in the OpenAPI spec.

POST https://ontoramp.com/api/checkout
Content-Type: application/json

{ "lookup_key": "governance_intelligence_bundle" }

# → { "url": "https://checkout.stripe.com/c/pay/..." }

See the full loop in worked tool calls on the agent-native walkthrough.