Documentation
Documentation
Everything you need to get the most out of yottacode, the command-line AI agent.
Getting started
Install on Linux or macOS, plus post-install setup.
Run your first coding session in minutes.
Configure providers, models, and behavior.
Core
Command-line interface and one-shot run mode.
Every built-in tool the agent can call.
Semantic navigation, diagnostics, and edits.
How yottacode remembers across sessions.
Delegate work to specialized agents.
How yottacode works under the hood.
Build, extend, and contribute.
Models & MCP
Choose and switch between models.
Connect Model Context Protocol servers.
Track token usage and billing per provider.
Providers
Required settings, profiles, image support, reasoning effort.
Native Messages API adapter for Claude models.
API keys, Responses API routing, hosted tools.
Sign in with a ChatGPT account (openai-auth).
Device-code auth against your Copilot plan.
Native Google generative-language adapter.
Gemini and Claude billed to your own GCP project.
Grok models with web and X search.
Local models, no API key required.
Any gateway speaking the OpenAI wire protocol.
100+ hosted models from build.nvidia.com.
Workflow
Manage and resume work sessions.
Parallel work with isolated git worktrees.
Issues, pull requests, and reviews.
Every TUI slash command, explained.
Customize the TUI appearance.
Opt-in experimental features.
Fan subtasks out to parallel subagents.
Cut screen recordings into publishable clips.
The media_* probe, compose, and render tools.
Generate xlsx, docx, pdf, and pptx files from structured content.
Run approved shell commands inside an experimental Podman sandbox.