Skip to content
robotadocs

Guide

Task-oriented guides for building agents with the Robota libraries and for using the robota CLI.

Contents

  • Architecture — The package layers, why they are separate, and how a request moves through them
  • Building Agents — The Robota agent class in agent-core: providers, tools, plugins, streaming, structured output
  • Using the SDK — agent-framework sessions: InteractiveSession, createQuery(), built-in tools, transports
  • CLI Reference — Every robota option, subcommand and slash command
  • Sessions, Background Sessions and the Daemon — Sessions that outlive the terminal, attaching, one daemon per workspace, workspace trust
  • TUI Keybindings — Contextual shortcuts, hot reload, chords, and terminal limits
  • Local LLM Setup — Ollama, LM Studio and llama.cpp with no API key
  • Providers — Every provider package, its options, and switching between them
  • Model Context Protocol (MCP) — Connecting to MCP servers, OAuth sign-in, robota mcp serve, external events
  • Embedding agent-framework — HTTP servers, bots, serverless functions and batch pipelines
  • Permissions and Hooks — Permission modes and rules, the OS sandbox, and lifecycle hooks
  • Devices, Peers and Remote Control — Messaging other sessions, linking your devices, handoff, co-driving from a browser
  • Context Management — Token tracking, compaction, and streaming
  • Error Handling — The typed error hierarchy and how to handle provider failures
  • Plugins — Runtime plugins for the Robota class and CLI plugins for robota
  • Deployment — One session served over many channels through the transport registry
  • Migrating to 3.0 — Package and API changes from the 2.x packages and between 3.0 betas