# PiCrust

## Docs

- [Context Injection](https://docs.framework.vibeworkapp.com/advanced/context-injection.md): Modify messages before each LLM call
- [Subagents](https://docs.framework.vibeworkapp.com/advanced/subagents.md): Spawn specialized agents from tools for complex multi-step tasks
- [AgentConfig API](https://docs.framework.vibeworkapp.com/api/agent-config.md): Configuration builder methods for agents
- [AgentHandle API](https://docs.framework.vibeworkapp.com/api/agent-handle.md): Methods for communicating with running agents
- [OutputChunk Reference](https://docs.framework.vibeworkapp.com/api/output-chunk.md): All output variants from agents
- [Agents](https://docs.framework.vibeworkapp.com/concepts/agents.md): Understanding the agent architecture in PiCrust
- [Architecture Overview](https://docs.framework.vibeworkapp.com/concepts/architecture.md): Understand the high-level architecture of PiCrust and how components work together
- [Model Context Protocol](https://docs.framework.vibeworkapp.com/concepts/mcp.md): Integrate MCP servers with PiCrust
- [Message Flow & Communication](https://docs.framework.vibeworkapp.com/concepts/message-flow.md): How messages flow between your application, agent handles, and agents - the critical subscribe-before-send pattern
- [Runtime & Agent Lifecycle](https://docs.framework.vibeworkapp.com/concepts/runtime.md): Deep dive into AgentRuntime and how agents are spawned, managed, and shut down
- [Sessions & Persistence](https://docs.framework.vibeworkapp.com/concepts/sessions.md): How conversation history and metadata are persisted to disk for durability and continuity
- [Agent States](https://docs.framework.vibeworkapp.com/concepts/states.md): Deep dive into all agent states, state transitions, and how to monitor agent status
- [Tools](https://docs.framework.vibeworkapp.com/concepts/tools.md): Built-in and custom tools for PiCrust
- [Features](https://docs.framework.vibeworkapp.com/features.md): Explore the powerful features of PiCrust
- [Ask User Questions](https://docs.framework.vibeworkapp.com/features/ask-user-questions.md): Interactive workflows with multiple-choice questions
- [File Attachments](https://docs.framework.vibeworkapp.com/features/attachments.md): Send files in user messages using vibe-work-attachment tags with automatic processing and deduplication
- [Extended Thinking](https://docs.framework.vibeworkapp.com/features/extended-thinking.md): Enable Claude to reason deeply about complex problems with extended thinking mode
- [Image & PDF Support](https://docs.framework.vibeworkapp.com/features/images-pdfs.md): Vision and document understanding with automatic base64 encoding
- [Interrupt Handling](https://docs.framework.vibeworkapp.com/features/interrupts.md): Graceful cancellation during streaming, permission waiting, and tool execution
- [Prompt Caching](https://docs.framework.vibeworkapp.com/features/prompt-caching.md): Reduce API costs by up to 90% with automatic prompt caching
- [Streaming & History](https://docs.framework.vibeworkapp.com/features/streaming.md): The dual-channel architecture pattern for real-time UIs with reliable persistence
- [Hooks](https://docs.framework.vibeworkapp.com/hooks/overview.md): Intercept and modify agent behavior at runtime
- [Tauri Integration](https://docs.framework.vibeworkapp.com/integration/tauri.md): Build desktop apps with PiCrust and Tauri
- [Introduction](https://docs.framework.vibeworkapp.com/introduction.md): A Rust framework for building AI agents with Claude, Gemini, and OpenAI - designed for applications that need autonomous, tool-using agents
- [LLM Providers](https://docs.framework.vibeworkapp.com/llm/providers.md): Pluggable LLM backends - Anthropic, Gemini, OpenAI, and custom providers
- [MCP Integration](https://docs.framework.vibeworkapp.com/mcp/integration.md): Connect to external MCP servers for dynamic tool discovery
- [Dangerous Skip Permissions Mode](https://docs.framework.vibeworkapp.com/permissions/dangerous-skip.md): Bypass permission checks for automation and testing
- [Handling Permission Requests](https://docs.framework.vibeworkapp.com/permissions/handling.md): Build permission UI flows for tool approval
- [Three-Tier Permission System](https://docs.framework.vibeworkapp.com/permissions/three-tier.md): Session, local, and global permission rules for controlling tool access
- [Quickstart](https://docs.framework.vibeworkapp.com/quickstart.md): Build your first Agent in under 5 minutes with this step-by-step guide
- [Built-in Tools](https://docs.framework.vibeworkapp.com/tools/built-in.md): Reference for all built-in tools - file operations, shell execution, and task management
- [Creating Custom Tools](https://docs.framework.vibeworkapp.com/tools/custom.md): Build custom tools to extend agent capabilities