Agents
TF Code ships with three built-in agents — Build, TF Engineer, and TF Orchestrate
MCP Servers
Configure MCP servers — ToothFairyAI MCP, local servers, OAuth, MCP Apps (interactive UI), skills, and functions
Tools
Built-in tools available to TF Code agents and how to manage them
Permissions
Control what actions TF Code agents can take
Compaction
Manage conversation context with automatic and on-demand compaction
Gateway
Webhook receiver gateway for connecting TF Code to external services like GitHub, Gitea, GitLab, and AWS Batch
GitHub Webhook Setup
Step-by-step guide to connecting TF Code to GitHub for automated ticket processing in a secure, production-grade deployment
Loops
Structured execution-path loops — named steps the agent traverses with multi-agent transitions, per-step compaction, reviewer, model_swap, and skills
Workspace Sync & Share
Move agents and skills between your local TF Code environment and a ToothFairyAI workspace
Scheduling
Recurring AI agent jobs and routines — schedule work on a cron schedule across the app, server, and TUI, with wake-and-decide gates, missed-run recovery, run history, and logs
Policies
Allow/deny policy statements that govern what TF Code agents may do — defined locally or enforced from your workspace
Escalate
Temporarily upgrade the active model to a stronger deployed model, matched on problem domain, then revert
Office documents
High-fidelity previews and desktop editing via LibreOffice, fully local
Kanban
A local kanban board as the shared coordination layer between agents and the human — boards, lanes, columns, tasks, comments, WIP limits, and an auto-running dispatcher
Channels
A Zulip-style conversation layer between the human and agents — high-level feedback, decisions, and insight logs, separate from task tracking
TF Design
Delegate design work — decks, posters, landing pages, prototypes, design systems, and brand visuals — to the local TF Design workspace