Guides

Task-oriented how-tos for specific goals. Pick the guide that matches what you want to do.

Getting started

• Run the Setup Wizard — Configure a provider, API key, and default model interactively

Models, providers & auth

• Configure Models & Auth — Connect OpenAI-compatible, Anthropic, and Gemini endpoints
• Token Caching — Reduce cost with prompt caching

Agentic features & harness

• Use Sub-Agents — Spawn specialized agents to parallelize work
• Use Agent Teams — Orchestrate multi-agent pipelines
• Use Skills — Package reusable instructions as installable skills
• Use Hooks — Run shell commands on agent lifecycle events
• Work Toward a Goal — Drive long-horizon tasks with goal mode

Tools, MCP & LSP

• Connect via MCP — Attach external tools via the Model Context Protocol
• Language Server Protocol — Give the agent live diagnostics and go-to-definition

Context & memory

• Manage Memory — Persist context across sessions
• Ignore Files — Exclude paths from the agent's file access
• Export Sessions — Save and replay conversation history

Permissions & safety

• Approval Mode — Control which tools run automatically vs. require confirmation
• Sandboxing — Isolate agent execution in a container
• Trusted Folders — Pre-approve tool use for specific directories

Running & automation

• Run Headless — Drive proto non-interactively from scripts and CI
• Schedule Prompts — Run prompts on a cron schedule
• GitHub Actions — Run proto in CI workflows

Editors & protocols

• ACP Coding Agent — Drive proto as a coding agent from an editor or orchestrator over the Agent Client Protocol
• Talk to A2A Agents — Use proto as a terminal client for protoAgent (and any A2A 1.0 agent): register, discover, and chat
• Zed — Use proto inside the Zed editor over ACP

Interface & customization

• Themes — Change the terminal UI color scheme
• Voice Input — Push-to-talk dictation with Ctrl+Space