Skip to content

PROJECT_SCOPE.md

Latest commit

 

History

History
82 lines (50 loc) · 3.83 KB

PROJECT_SCOPE.md

File metadata and controls

82 lines (50 loc) · 3.83 KB

Project Scope

Agentage Desktop — Simplify your agent management.

Agentage Desktop gives you one place to discover, run, and manage all your AI agents. Instead of juggling terminals and tracking runs in your head, you get a clean dashboard, automatic agent discovery, real-time streaming, and full run history — across every project you work on.

Core Identity

Agentage Desktop is an AI agent management app. It takes the friction out of working with agents:

  1. Run agents easily — Select an agent, provide inputs, and run. Agents stream output in real time so you can watch progress as it happens. No terminal juggling.
  2. Approve and interact — When an agent needs your input, it pauses and waits. Approve, reject, or guide it — then it picks up right where it left off.
  3. Task dashboard — A single view of all running and completed agents across all your projects. No more hunting through terminal tabs or forgotten sessions.
  4. Agent discovery — Drop an agent definition in the right place and Agentage finds it automatically. Browse everything in one catalog.
  5. Run history and tracking — Every run is logged with full output, token usage, and timing. Review what happened, compare runs, and track costs.

What's Being Built Now

Task dashboard redesign — A unified dashboard to track agent runs, review results, and manage tasks at a glance across all projects.

Agent catalog improvements — Better browsing, filtering, and organization for discovered agents.

Agent Execution

When you run an agent, it goes through a simple lifecycle:

submitted → working → completed / failed / canceled

Agents can also pause with input_required — the agent stops and waits for your response before continuing. While running, agents stream output in real time. Run history and token usage are tracked automatically.

Agent Discovery

Agentage automatically discovers agents from your filesystem — no manual registration needed. You can define agents as simple markdown files or as TypeScript code for full control. The discovery system is evolving and may change, but the principle stays the same: drop a file, see it in the app.

Supporting Features

Feature What it does
Project management Register local repositories, manage settings, organize work
Git integration Status, branches, diffs, worktrees
AI chat Conversational AI interface (desktop mode)
Task boards Kanban-style task organization within projects
Inbox Issue triage from GitHub, Sentry, and other sources
Cross-device access Web mode lets you access agents from other devices (planned)

What Agentage is NOT

  • Not an IDE — It runs agents, not edits code
  • Not just a chat app — Chat is a supporting feature; the core is agent management
  • Not a CI/CD tool — It manages agent runs, not deployments
  • Not a framework — It's an end-user app, not an SDK

Target Users

Developers and teams who:

  • Run AI agents regularly and want a simpler way to manage them
  • Work across multiple projects and need one place to see all their agents
  • Want automatic agent discovery instead of manual setup and configuration
  • Need run history, token tracking, and real-time streaming for agent executions

Architecture (for contributors)

Electron + React + Express monorepo with TypeScript. Runs as a desktop app (Electron/IPC) or web app (HTTP/Express) from the same codebase.

See docs/ARCHITECTURE.md for the full technical breakdown.