Introduction

Introduction to WorkPilot AI

What is WorkPilot AI?

WorkPilot AI is a desktop application (also available as a CLI) that lets you describe a software goal in natural language — and lets AI agents handle the entire development cycle: planning, implementation, testing, quality review, then merge.

Instead of copying AI-generated code into your editor, WorkPilot AI:

1. Analyzes your project automatically (framework, language, test tooling…)
2. Creates a detailed spec from your request
3. Breaks the work down into ordered subtasks
4. Codes in an isolated workspace (git worktree)
5. Validates with a QA agent that checks acceptance criteria
6. Fixes automatically anything that fails
7. Proposes a semantic merge back to your main branch

All of this without ever touching your main branch until you approve.

---

Who is WorkPilot AI for?

🎯 Individual developers

• Prototype features faster
• Automate repetitive tasks (refactoring, migrations, tests)
• Explore an unfamiliar codebase with the Insights agent

🏢 Teams and enterprises

• Automate triage and resolution of GitHub/GitLab/Jira issues
• AI-assisted PR reviews
• Test and documentation generation at scale
• Automated incident response (Self-Healing)

🎓 Learners and curious minds

• Learning Mode — understand agent decisions step by step
• Pair Programming — code side-by-side with an AI
• Arena Mode — compare multiple AI models on the same task

---

How is it different from other AI tools?

Typical tool (Copilot, Cursor…)	WorkPilot AI
Inline autocomplete or chat	Autonomous agents that work end to end
One model at a time	Multi-provider + multi-account, auto-switches on rate limit
You drive every step	Autonomous pipeline: Planner → Coder → QA → Fixer
Modifies your files directly	Isolated worktrees — your main stays untouched
No global view	Visual Kanban + Mission Control to orchestrate many agents
No cross-session memory	Graphiti — persistent graph memory

---

Flagship features

🔄 Autonomous development pipeline

Every task flows through a full pipeline: a Planner assesses complexity, a Coder implements (with parallel subagents if needed), a QA Reviewer validates acceptance criteria, and a QA Fixer automatically corrects issues for up to 50 iterations.

🎛 Multi-agent parallelism

Up to 12 simultaneous agent terminals. Orchestrate multiple builds in parallel from Mission Control — a NASA-style dashboard showing real-time status, token consumption, file changes, and per-agent reasoning.

🏝 Isolated workspaces

Each task runs in a dedicated git worktree. Your main branch is never modified until you approve the merge. An AI-powered semantic merge engine resolves conflicts when multiple worktrees come back to main.

🔐 Three-layer security

1. OS Sandbox — bash commands run in isolation
2. Filesystem restrictions — operations limited to the project directory
3. Dynamic allowlist — only commands approved for the detected stack are allowed

🌍 Multilingual, multi-theme

Full French and English UI, 7+ color themes (Default, Dusk, Lime, Ocean, Retro, Neo…) each with light and dark variants.

🧠 Persistent memory (Graphiti)

Agents retain insights across sessions via a semantic graph. They no longer need to rediscover your architecture on every task.

🔌 Rich integrations

GitHub, GitLab, Azure DevOps, Linear, Jira, Windsurf, MCP Marketplace. Import issues, let AI investigate them, auto-create PRs.

---

Concrete use cases

Situation	What WorkPilot AI does
"Add OAuth authentication to my API"	Plans, codes, tests, documents — all in an isolated worktree
"Triage the 50 open GitHub issues"	Imports, categorizes by severity, proposes investigations
"Migrate this project from React 17 → 19"	Uses the Migration Agent with breaking-change detection
"Why does this test fail in CI?"	Self-Healing analyzes the log, proposes and applies a fix
"Explain this codebase to me"	The Insights agent answers with semantic search
"Compare Claude vs GPT-4 on this refactor"	Arena Mode runs the same task on multiple models

---

Next step

➡️ Installation & Quick Start