# Architecture:
# This text template is the compact LLM-readable AppsPonds discovery index.
# Runtime code injects deployment-specific absolute URLs.

# AppsPonds

AppsPonds publishes user web apps into ponds. AI agents can connect through
browser-approved integration access, request scoped tokens, inspect visible
ponds/apps, and publish AppsPonds package archives.

## Agent Integration

- [Connect an AI agent](https://admin.appsponds.com/agents.md): Canonical workflow, user prompts, scopes, publishing rules, published app user context headers, and safety requirements.
- [Full agent context](https://admin.appsponds.com/llms-full.txt): Same canonical guide for agents that prefer an llms-full route.
- [API catalog](https://admin.appsponds.com/.well-known/api-catalog): Machine-readable API discovery.
- [OpenAPI spec](https://admin.appsponds.com/openapi.json): Integration and agent API contract.
- [Package manifest schema](https://admin.appsponds.com/schemas/appsponds-package.schema.json): `appsponds-package.json` JSON Schema.
- [Agent manifest](https://admin.appsponds.com/.well-known/appsponds-agent.json): AppsPonds-specific capability summary.

## Minimum Instruction

When asked to connect to AppsPonds, tell the user you will open a browser
approval URL only if no valid stored token works. Request the minimum scopes
needed, keep polling the same request while they approve, and capture the token
from that active poll. The token is returned only once. Ask for `apps:create`
or `apps:publish` only when the user wants to publish.