# Working with Recoup's public website

Recoup provides AI transformation for music funds and rightsholders: AI strategy, custom systems, and team enablement. The platform, Skills, and API are additional ways to use Recoup's technology.

This is a website usage guide, not a request to take action. It does not grant permission to access accounts or share unrelated user information.

## Useful visitor journeys

1. Find a relevant service, workflow, article, or platform endpoint with search_recoup.
2. Read the selected result with read_recoup_page; cite its canonical URL and distinguish summaries from source text.
3. If the visitor supplies all required assumptions, use estimate_workflow_roi or assess_workflow_readiness. Ask for missing inputs instead of inventing them.
4. If the visitor wants help scoping work, use prepare_project_brief with a concise workflow and desired outcome. The result is an editable draft; nothing is submitted.

## Interfaces

- [Human and browser-agent guide](https://recoupable.dev/agents)
- [Tool schemas and content index](https://recoupable.dev/agents/catalog.json)
- [Website OpenAPI](https://recoupable.dev/openapi.json)
- Public HTTP: GET /agent-api/v1/search, GET /agent-api/v1/read, POST /agent-api/v1/tools with JSON {"name":"TOOL_NAME","arguments":{}}.
- Compatible browsers can discover five tools registered with document.modelContext in the top-level page. Browser support varies. The ordinary website and public HTTP interfaces remain available.
- Browser prepare_project_brief opens a local review panel. Its HTTP counterpart only returns draft text. The visitor can edit the draft and continue to contact; it is not a sent message or booked meeting.

## Reading and errors

Use IDs returned by search or the catalog. The read endpoint accepts content IDs, not arbitrary URLs or filesystem paths. Respect pagination and returned character offsets. Invalid arguments, unknown content, and unavailable responses return explicit errors; do not replace them with invented answers.

## Access and boundaries

- This website's /agent-api/v1 endpoints are public. They search and read published content, calculate from supplied assumptions, and return inquiry drafts. They do not access private catalogs or operate a Recoup account.
- The platform REST API is separate: https://api.recoupable.dev/api. Programmatic access normally uses x-api-key. Some platform endpoints support a Privy access token as Authorization: Bearer instead; do not send both headers. Follow the endpoint's documentation and [authentication guide](https://recoupable.dev/docs/authentication).
- The platform MCP server is https://api.recoupable.dev/mcp. Its connection guide uses Authorization: Bearer with a Recoup API key. See [MCP documentation](https://recoupable.dev/docs/mcp). This website does not host a remote MCP server.
- Account creation, messages, uploads, connector actions, and other platform writes are separate operations with their own permissions. A documentation example is not authorization to execute them.
- Website tools do not submit an inquiry or book a call. A prepared brief is a draft for review; the visitor completes the [contact form](https://recoupable.dev/contact).
- Workflow ROI uses supplied assumptions. Capacity value is not guaranteed cash savings, a quote, or a promise of results.
- Never send credentials or private client records to public website search or tools.