# XSAF

A small, fetch-native TypeScript framework for building secure AI agents.

Full corpus (all pages, one document): https://xsaf.ilha.build/llms-full.txt

## Pages

- [Builder and lifecycle](https://xsaf.ilha.build/components/index.md) — Configure, seal, start, invoke, and stop an XSAF agent.
- [Delegation](https://xsaf.ilha.build/delegation/index.md) — Delegate focused tasks to sealed child agents with explicit context, approval, and sandbox boundaries.
- [Getting started](https://xsaf.ilha.build/getting-started/index.md) — Install XSAF and run a deterministic agent without API tokens or network calls.
- [MCP](https://xsaf.ilha.build/mcp/index.md) — Serve tools and consume remote tools with MCP protocol version 2026-07-28 over HTTP.
- [API reference](https://xsaf.ilha.build/reference/index.md) — Published imports, public methods, structural contracts, and events in XSAF 0.1.0-alpha.0.
- [Roadmap](https://xsaf.ilha.build/roadmap/index.md) — Current alpha boundaries, security caveats, and features planned beyond 0.1.0-alpha.0.
- [Scheduling](https://xsaf.ilha.build/scheduling/index.md) — Run prompts through the normal agent path with process-local five-field cron schedules.
- [Sessions & Memory](https://xsaf.ilha.build/sessions-memory/index.md) — Understand session isolation, streaming completion, memory persistence, channels, and the fetch-native runtime.
- [Testing](https://xsaf.ilha.build/testing/index.md) — Test agents, channels, streaming, tools, and Hono routes offline with deterministic mocks.
- [Tools & Security](https://xsaf.ilha.build/tools/index.md) — Define tools, validate inputs, control approval and retries, and require an explicit execution boundary.
- [Terminal UI](https://xsaf.ilha.build/tui/index.md) — Add a polished interactive prompt, Markdown history, streaming output, and runtime statuses with @xsaf/tui.