BTALabs

Guide

What is an agent development environment (ADE)?

An agent development environment is the software you open when the work is handing tasks to coding agents rather than typing code yourself. The editor is where you write; the ADE is where the agents you hired do.

The definition

An agent development environment (ADE) is a workspace built around coding-agent CLIs. It hosts them in real terminals, gives each one its own checkout so parallel work cannot collide, and keeps the result — a diff, a branch, a merge decision — in front of you.

The idea it is organised around is a hand-off: you describe a task, an agent does it somewhere that is not your working copy, and you decide whether the result lands. That is the whole shape. Everything else is detail about how well it is done.

  • Terminals. Real PTY sessions, because the agents are CLIs and the escape hatch has to stay open.
  • Isolation. A git worktree per agent session — separate branch, separate working copy, shared object database.
  • Review. A diff per worktree and a merge queue, so nothing lands unreviewed.
  • Context. The project, its instructions, its sessions and their state, in one window that survives restarts.

What an ADE is not

It is not an IDE with a chat panel. An editor optimises the loop where you and a model edit one file together; an ADE optimises the loop where you hand work off and review it later. Editors are adding agents, which is fine — the difference is what the window is organised around.

It is not a terminal emulator. Ghostty, WezTerm, Windows Terminal and Warp make one shell pleasant to use. An ADE is what sits above them when the things in those tabs are agents that need separate checkouts and a review step. You keep the terminal; the ADE takes over the bookkeeping.

It is not a chat app. A chat answer is a suggestion. An agent in a worktree produces a diff you can accept, reject or send back — and that difference is the point.

And it is not the Nix package manager. Nix Agents is not related to Nix, NixOS or nixpkgs, is not a package manager, and does not use the Nix build system. The name here refers to this app, nothing else.

What to look for in one

A short checklist, in the order that matters:

  • Real terminals, including on your platform. On Windows, that means PTY sessions and not an emulation layer — the difference shows up the first time an agent draws a full-screen TUI.
  • One worktree per session, created and named automatically. If you are making the worktrees by hand, you have a terminal and a habit, not an environment.
  • A diff you can read in the app, and a merge you can decline without touching your working copy.
  • The agent CLIs you already use and already pay for — Claude Code, Codex, Gemini CLI and the rest — running as themselves, with your logins, not proxied through someone else’s API.
  • Local-first by default: your repository, your machine, your subscriptions.

Nix Agents, in one paragraph

Nix Agents is an ADE for Windows, in beta. It runs Claude Code, Codex, Gemini CLI, GitHub Copilot CLI and about thirty more in real terminals, each session in its own git worktree, on a board with five modes — Agent, Code, Chat, Loop and Swarm — plus source-control links, a project vault and SSH sessions for remote machines.

The Windows x64 and ARM64 installers are on the download page; the free tier covers chat and one code session, and the paid tiers raise the number of sessions and terminals.

Questions

What does ADE stand for?
Agent development environment — the workspace where coding agents are run, reviewed and merged, as opposed to the editor where you write code yourself.
Is an ADE the same as a terminal app?
No. A terminal app draws a shell for you. An ADE hosts terminals for the agents and adds the parts a terminal cannot express: a worktree per session, a diff per worktree, and a merge decision.
Is Nix Agents related to the Nix package manager?
No. Nix Agents is an agent development environment built by BTA Labs. It has no connection to Nix, NixOS or nixpkgs, is not a package manager, and does not use the Nix build system.
Do I need API keys to use an ADE?
Not for the ones that run CLIs as themselves. Nix Agents launches the agents you installed and never asks for provider credentials — the CLIs authenticate exactly as they do in your shell.
Do I need an ADE for one agent?
No. One agent in one terminal is fine. The arithmetic changes at two or three: worktrees, branch names, diffs and a merge order stop fitting in your head.
← All guides