PayWeavePayWeaveBack to Home
Discovery

Discovery Overview

PayWeave makes every published gateway, function, and file collection automatically discoverable to AI agents, LLMs, and developer tools through a set of open standard protocols. No extra configuration is required — enabling the Discoverable toggle on a resource is enough to make it appear across all discovery endpoints.

The three skill types

Discovery surfaces three kinds of monetizable resources, all unified under the Skill model:

  • Gateway — a reverse proxy to an external API. Each endpoint has its own path, HTTP method, and per-call price.
  • Function — a user-deployed serverless handler. Each handler has its own path and TypeScript code.
  • File collection — a group of monetized file downloads. Each file is a GET endpoint with its own price.

Discoverable flag

A skill only appears in discovery results when both conditions are true:

  • The resource's discoverable field is set to true
  • The resource's status is active

Paused or archived resources are hidden from all discovery endpoints even if discoverable is enabled.

Workspace scoping

Every discovery endpoint has both a global form and a workspace-scoped form. Global endpoints return resources from all workspaces that have chosen to make them discoverable. Workspace endpoints return only resources belonging to a single workspace, identified by its UUID.

Caching

All discovery responses are cached in a distributed KV store with a five-minute TTL. Changes to resource configuration (name, description, discoverable flag, price) may take up to five minutes to propagate to discovery endpoints.

Keep your resource names and descriptions clear and specific. These fields are what AI agents read when deciding whether a skill is relevant to their task.

Discovery protocols

PayWeave exposes four complementary discovery interfaces:

  • Skills API — a paginated JSON REST API for programmatic discovery with filtering and search.
  • LLMs.txt — a plain-text markdown file optimised for LLM consumption, listing all skills with pricing.
  • MCP Server — a Model Context Protocol server exposing list_resources, find_resources, and get_resource tools for agent tool use.
  • AI Plugin — an ai-plugin.json manifest for OpenAI-style plugin discovery.