Maestrly

Features overview

Maestrly is more than a launcher for AI tools. Most of the power lives in the conversation drawer — a set of surfaces attached to each conversation — plus an advanced toolkit you can reach for as your workflow grows. This page tours both.

The conversation drawer

Every conversation has a drawer (the button to the right of the conversation header). Each surface is scoped to that conversation, and the drawer is kept alive per conversation — so you can switch conversations without losing your terminals, browser tabs or editor state. Surfaces can sit inline, open as a tab, or be detached into their own window.

SurfaceWhat it's for
CodeA built-in editor to inspect and edit the files the agent is touching, without leaving the app.
TerminalBuilt-in terminals for the conversation. You can run commands yourself, and the agent can drive them too.
BrowserA built-in browser the agent can open and read — useful for docs, local previews and looking things up.
PlanReview, edit, comment on and approve the agent's plan before it executes — you stay in control of what happens.
ReviewSee the diff, pull request and checks for the conversation's branch, so you can approve changes before they ship.
NotesThe conversation and project notes (see Concepts) — readable and writable by you and the agents.
DelegationsHand a focused sub-task to another AI tool from inside the conversation, and watch it work in the drawer.

Advanced

These features go beyond the essentials. Reach for them as you take on bigger work.

  • Multi-CLI delegation — a "manager" conversation can delegate a focused task (confront a plan, review code, implement a change) to a different AI tool. The delegate runs visibly in the drawer and reports back, so you combine the strengths of several tools in one thread.
  • Debug from the IDE — the agents can drive a full debugger: set breakpoints, step through code, and inspect variables, controlled from within Maestrly.
  • Project memory — durable, per-project memory injected into the AI tools at startup, so important facts survive across conversations (see Concepts).
  • Built-in browser — beyond manual browsing, the agent can navigate and read pages to gather context as it works.
  • Multi-repo — a single conversation can touch several repositories at once, each isolated in its own worktree and branch, and reviewed per repository.
  • Workspace groups — organize the sidebar into virtual groups to keep many projects tidy. It's a view only — nothing moves on disk.
  • Project assistant — a management chat for the board: it organizes cards, comments and triages (it manages, and delegates the actual coding to card conversations).
  • Board automation — columns that automatically dispatch agents as cards move, with guardrails (anti-loop limits and per-board controls) so a runaway agent can't burn through your subscription.

Where to go next