scrollport gives an agent one authenticated way to discover, inspect, pay for and run catalog tools. This documentation explains the platform surface; the live catalog remains at /tools, where current categories, capabilities, providers and catalog tools are sourced from the public catalog.
Start with the first success
If you are setting up a new agent, begin with the quickstart. It takes you from the recommended setup prompt through device authorization, a wallet check and the four-control-tool loop:
discover → inspect → run → wallet
The sequence is deliberately small. Catalog tools are discovered on demand; they are not a second set of permanently loaded MCP definitions.
Choose a path
- Connect your agent explains the browser-approved device flow.
- How scrollport works explains the catalog, provider boundary, wallet and run lifecycle.
- Control tools gives the contract for each of the four agent-facing operations.
- Guides covers first runs, approvals, wallet behavior and recovery.
- Reference records transport, authentication and error details.
Keep the source of truth in the right place
Documentation describes stable platform behavior and teaches an agent how to use it. Provider-specific schemas, prices, connection state, availability and worked examples belong to the live catalog and inspect(tool_id). Do not copy those mutable facts into documentation.
Explore the current catalog when you need a provider-backed tool, then inspect the exact tool_id returned by discover before running it.