Isolated By Default
Each session runs in its own worktree and branch, so experiments stay contained.
// terminal-first agent workflows
Agentty is an ADE (Agentic Development Environment) for structured, controllable AI-assisted software development. This project is developed with Agentty itself, maintaining the same AI-assisted workflow it provides.

npm install -g agentty// designed for control
Each session runs in its own worktree and branch, so experiments stay contained.
Review diffs, switch models, and iterate without leaving the TUI.
Open PRs or MRs, track status, and auto-resolve threads after agents fix them.
Chain dependent sessions; Agentty automatically rebases review-ready children.
Each changed turn refreshes one AI-written commit and updates its published branch.
On conflicts, the session agent resolves files and Agentty continues the rebase.
Track sessions across models with color-coded statuses. Active, queued, and archived work at a glance.

// roadmap
Where Agentty is today, and the product paths growing from it.
The first version of Agentty as an orchestrator: a terminal-first environment for running coding agents in isolated workspaces and steering changes from prompt to review.
Provide a lightweight integration point for LLMs, tools, context, and model providers without requiring the full Agentty interface.
Coordinate specialized agents as one system: split work, run tasks in parallel, manage dependencies, and keep handoffs visible.
Review pull requests across open-source GitHub projects, surface actionable findings, and help maintainers assess changes at scale.
Explore a user-controlled assistant that carries context across tasks, invokes tools, and helps across projects.
Run durable agent sessions as a managed service with remote execution, shared environments, and team visibility across devices.
Strategic directions, not release commitments.