AI DEV CONTROL CENTER

Claude develops. Codex reviews. AutoPilot manages.

AutoPilot turns AI coding into a controlled development workflow. Claude builds, Codex reviews, and AutoPilot manages tickets, approvals, logs, memory, and delivery — from one professional control center.

AutoPilot — Working View
Develops
Claude
Reviews
Codex
Plan approved
Code implemented
Review & tests running
Claude develops — Codex reviews — AutoPilot manages — you stay in control
The Workflow

Claude develops. Codex reviews. AutoPilot manages.

One control center coordinating the full AI development workflow — with you approving every important decision.

01Claude DevelopsClaude writes and implements the code for each ticket — new features, fixes, refactors, and tests — working from a clear, scoped task.
02Codex ReviewsCodex reviews and validates every change, catches issues, and sends corrections back through the loop before anything is marked done.
03AutoPilot ManagesAutoPilot runs the whole operation — projects, tickets, approvals, workflow, memory, logs, and execution — keeping every run on track.
The Problem

AI coding is powerful. The workflow is still manual.

You lose hours moving prompts between tools, rebuilding lost context, and running the same review loop again and again.

📋

Copy-paste tax

Shuttling prompts between Codex and Claude by hand.

🧠

Context loss

Tools forget the project between tasks.

🔄

Manual review loops

You broker every fix, retry, and re-check.

🔔

Constant interruptions

Approvals and questions break your flow.

💸

Wasted tokens

The same context re-sent on every call.

🛡

No safe automation

No guardrails to let agents run unattended.

The Solution

AutoPilot manages the AI development loop for you.

Hand AutoPilot a task list. It plans the work, delegates to Claude, reviews with Codex, then compiles, tests, and retries until it passes.

Claude develops — writes and implements the code
Codex reviews — validates, catches issues, corrects
AutoPilot manages — tickets, approvals, logs, memory, execution
You stay in control — approve the final result
AutoPilot development dashboard showing Claude and Codex dev loop
AutoPilot manages the development loop while Claude builds and Codex reviews.
How It Works

From task to working code in 8 steps.

1Connect Codex and ClaudeLink your local AI tools to AutoPilot.
2Create or scan a projectStart fresh or import an existing codebase.
3Add your tasksTickets, bugs, epics, or refactoring tasks.
4AutoPilot plans the workBreaks the task into steps and queues it for the agents.
5Claude developsClaude writes and implements the code.
6Codex reviewsCodex validates the change, corrects, and gives feedback.
7Retry, compile, testAutoPilot handles the build cycle.
8You approveReceive a ready-to-test version.
Features

Everything you need for AI-powered development.

Auto Resume on Rate LimitsAutoPilot detects rate limits and automatically resumes work when the limit resets. No manual intervention needed.
📚Multi-Project WorkspaceWork on multiple projects simultaneously with separate memory and task history for each one.
🌐Multi-Language SupportBuild applications in any language. AutoPilot adapts its workflow to your project’s stack.
🧠Task MemoryAutoPilot maintains its own memory for each task, so the next round knows what was already done.
🔍UI ReviewScreenshot review workflow lets you visually verify changes before approval.
🔗Third-Party IntegrationsConnect with your existing development tools and workflows.
🛠Compile and Test WorkflowAutoPilot compiles code and runs tests as part of the development loop. Failures trigger automatic retries.
📈Clear Status and DetailsA simplified development process with clear status updates, logs, and progress details at every step.
🛡Advanced PermissionsAllow All mode with warnings for experienced developers who want maximum speed.
What’s New

Just shipped: a major update.

Built-in Git management with safe commit, push, and one-click version history; an “Ask a Question” chat that knows your whole codebase; optional Plan Mode for features; a more resilient autonomous worker; and automatic in-app updates.

NEW🌿Git Project ManagerUnderstand and safely manage Git for every project. Live repo status in the top bar, a full Git Manager (status, commits, branches, remotes), and commit/push without leaving the app. Turn on Auto-Commit after each approved ticket. Every destructive action is guarded behind an explicit confirmation.“Version control, built in. Nothing risky without your say-so.”
NEW💬Ask a QuestionA ChatGPT-style chat that knows your codebase. Codex answers first with full repo context and brings in Claude automatically for deeper reasoning. Rich markdown answers, screenshot attachments, multi-turn history — and minimize it to keep running in the background while you work.“Ask your codebase anything — without stopping your work.”
NEW🗺Plan Mode for ticketsFeature tickets can run a plan-first workflow: Claude drafts a structured implementation plan, open questions get resolved automatically by Codex or by you in a review dialog, and nothing is built until the plan is approved.“Think before you build — optional planning for bigger features.”
NEW🔁A worker that never gets stuckThe autonomous maintenance worker keeps moving. If a ticket stalls, it verifies whether the work is actually done and either completes it or moves on — and when it genuinely needs a decision, it asks you in a dialog instead of waiting silently.“Autonomous that stays autonomous.”
NEXT RELEASE🔄Automatic updatesAutoPilot is becoming a proper Windows installer that updates itself. A non-intrusive banner notifies you when a new version is available; it downloads only after you approve and never restarts mid-task — installing once your current task finishes.“Always up to date, never interrupted.”
AutoPilot existing codebase scan showing project structure and Codex understanding
AutoPilot maps your codebase without modifying source files.
Existing Projects

Bring AutoPilot into an existing project.

AutoPilot scans your codebase, detects the stack, maps the key files, and builds a project profile — so you can ship features right away.

Detects languages, frameworks, and tools
Identifies key files and entry points
Summarizes architecture and patterns
Builds a project profile for agents
Automation Control

You decide how automatic it should be.

Set your automation level from fully manual to safe autopilot.

0% — Ask me everything50% — Codex decides100% — Run safely

Safety at every level: At high automation, AutoPilot auto-approves safe, project-local actions and keeps moving. Dangerous actions — secrets, publishing, system changes, anything outside the project — stay blocked or need your explicit approval.

Token Saver Mode

Designed to reduce unnecessary token usage.

AutoPilot trims token waste with compact JSON prompts, compressed memory, slim logs, and tightly focused context.

In long Codex–Claude runs, that can cut unnecessary context by roughly 40–60%, depending on task size and history.

Compact structured JSON prompts
Compressed task memory
Trimmed logs
Focused context windows
~40-60%
reduction in unnecessary token usage
in many agent-loop workflows
Standard Mode100%
Token Saver Mode~45%
AutoPilot agile board managing development tickets with Claude and Codex
AutoPilot turns AI development into a managed delivery workflow.
Task Management

AI development needs a real task system.

Track features, bugs, epics, refactors, docs, and tests. AutoPilot keeps history per task, so every round knows what already shipped.

FeaturesBugsEpicsRefactoringTestsDocsResearch
See It In Action

See AutoPilot in action

Claude develops. Codex reviews. AutoPilot manages the workflow.

What the demo shows: A real run inside AutoPilot, the AI Dev Control Center. A task is added to the board, Claude develops the change, Codex reviews and validates it, and AutoPilot manages the loop — tracking tickets, approvals, logs, and memory while it compiles, tests, and retries until the work passes. You approve the result and stay in control the entire time.

Privacy

Local-first. Private by default.

AutoPilot works with your local project folders and local AI coding tools. Your source code, prompts, logs, task history, and agent conversations stay on your machine by default. AutoPilot does not upload your codebase to our servers. Pro license checks and app updates only use minimal account, license, app version, and entitlement metadata.

Your code stays local.
AutoPilot is local-first.
No source code is uploaded to our servers by default.
License checks only use minimal license and account metadata.
🔒

Stays on your machineSource code, prompts, logs, task history, and agent conversations never leave your device by default — only minimal license and update metadata is exchanged.

Pricing

Get started for free. Scale when ready.

Free
$0
No account required
Free desktop app download
100% automation
One open project & worker loop
Plan Mode + Claude/Codex loop
Codebase scan & local memory
Basic Git Manager & Agile Board
RECOMMENDED
AutoPilot Pro
Pro
Advanced workflows & multi-project work
Multiple open projects / tabs
Multi-project parallel work
Full Git Manager & commit timeline
Auto Commit after approved ticket
Advanced Token Saver & Auto-Resume
Export reports & advanced history
Priority beta features
Power / Team
Coming Soon
For teams and organizations
Auto Push & Allow All Mode
Telegram / Slack integrations
GitHub / Bitbucket / Jira
Team workspace & shared policies
Audit logs & admin controls
Coming Soon
How Pro licensing works

AutoPilot Free works without registration. When you upgrade to Pro, Lemon Squeezy securely handles the checkout and sends you a license key. Open AutoPilot, choose Enter License Key, paste your key, and Pro unlocks on your device.

See full pricing & licensing →

FAQ

Frequently Asked Questions

AutoPilot is an AI Dev Control Center. It manages the full AI development workflow: Claude develops the code, Codex reviews and validates it, and AutoPilot manages projects, tickets, approvals, workflow, memory, logs, and execution — while you stay in control of the process.
No. AutoPilot is not a replacement for Claude or Codex. It manages the workflow around them — turning your AI coding tools into a controlled, repeatable development process with tickets, reviews, approvals, memory, and logs.
No. AutoPilot keeps developers in control. It coordinates AI agents and automates the busywork, but you decide what gets built, review the work, and approve every important step.
Claude is the developer in the loop. AutoPilot hands Claude clear, scoped tickets, Claude writes and implements the code, and AutoPilot tracks the result, memory, and logs across the whole project.
Codex is the reviewer. After Claude develops a change, Codex reviews and validates it, flags issues, and sends corrections back through the loop before anything is marked done.
Yes. AutoPilot can scan an existing project, detect the stack, map the key files, and build a project profile — so you can start managing features, fixes, and refactors in an existing codebase right away.
Yes. AutoPilot is local-first. It works directly with your local project folders and local AI coding tools, and your source code stays on your machine by default.
AutoPilot is local-first today. Remote and team workspaces are on the roadmap — check the roadmap page for what’s planned next.
AutoPilot runs locally on your machine. Your code stays on your computer. The AI models (Codex and Claude) are accessed through their respective APIs, but AutoPilot itself does not require a cloud server or remote service to function.
AutoPilot uses both the OpenAI API (for Codex) and the Anthropic API (for Claude). You connect your own API keys. AutoPilot does not proxy your requests or store your API keys on any external server.
No. AutoPilot never asks for or stores your OpenAI or Anthropic account passwords. You provide API keys, which are stored locally on your machine and are never transmitted to our servers.
Yes. AutoPilot can scan your existing codebase, detect the stack, identify important files, summarize the architecture, and build a project profile. You can start adding features to an existing project immediately after the scan.
Yes. AutoPilot supports a multi-project workspace. Each project has its own memory, task history, and configuration. You can switch between projects at any time.
It can be, but you control how automatic it is. The automation slider lets you choose between fully manual (you approve every step), semi-automatic (Codex decides routine tasks), and safe autopilot (runs until done, blocking only dangerous actions).
Safety is built into every automation level. AutoPilot blocks dangerous actions like publishing, deleting files outside the project, accessing secrets, and modifying system files. At higher automation levels, only safe, project-local operations are approved automatically.
Yes. You can pause, review, modify, or cancel any task at any time. AutoPilot is designed so that you remain in control, even when running in autopilot mode.
Token Saver Mode is designed to reduce unnecessary token usage by using compact structured JSON prompts, compressed task memory, trimmed logs, and focused context windows. In many agent-loop workflows, this can reduce unnecessary Claude-call context by roughly 40–60%, depending on task size, logs, history, and prompt mode.
AutoPilot is for developers who already use AI coding tools like Claude and Codex and want to connect them into a managed, repeatable development workflow. It is built for solo developers, freelancers, and small teams who want to move faster without losing control.
No. AutoPilot Free can be downloaded and used without creating an account. A Pro subscription / license key is only required for Pro features.
Buy AutoPilot Pro through the website. Lemon Squeezy will provide a license key after purchase. Open AutoPilot, choose Enter License Key, and paste your license key to unlock Pro features.
After purchasing AutoPilot Pro, Lemon Squeezy provides your license key in the order details and email receipt. You can also access your order through Lemon Squeezy’s customer order page.
Yes. AutoPilot Pro is a subscription handled by Lemon Squeezy. You can manage or cancel your subscription through the Lemon Squeezy order / customer portal.
No. AutoPilot is local-first. Your source code, project folders, prompts, logs, task history, and agent conversations stay on your machine by default. AutoPilot does not upload your codebase to our servers for normal local usage.
Pro licensing may use minimal metadata such as license key status, app version, license activation, and entitlement status. It does not require sending your source code, prompts, logs, or project conversations.
The Free version includes the core AutoPilot experience: 100% automation, one open project at a time, Plan Mode, Claude and Codex workflow loop, Ask a Question, codebase scan, local memory, UI screenshot review, basic Git Manager, manual commit, and auto-updates.
Pro unlocks advanced workflows such as multiple open projects, multi-project parallel work, full Git Manager, Auto Commit, advanced Token Saver, advanced auto-resume, export reports, and advanced project history.
No. AutoPilot is a self-serve desktop software product for developers. AutoPilot does not sell custom software development services, IT consulting, or agency work.
Who It’s For

Built for everyone shipping with AI coding tools.

If you already build with Claude and Codex, AutoPilot gives you the control layer your workflow has been missing.

👨‍💻Solo developersDevelopers using Claude, Codex, and other AI coding tools who want a repeatable workflow instead of endless copy-paste.
🚀Startup foundersFounders shipping MVPs fast who need real delivery and momentum — not just isolated code snippets.
🏢AgenciesTeams managing multiple client projects, each with its own board, memory, history, and approval flow.
📊Product managersPMs coordinating AI development with clear tickets, real-time status, logs, and explicit approvals.
👥Engineering teamsTeams already using Claude Code, Codex, Git, and issue workflows who want them connected into one controlled loop.
🧑‍💻Freelancers & consultantsIndependent builders delivering client work with AI — on time, on scope, and under control.
Why Now

AI coding tools changed development. The workflow needs to catch up.

Claude and Codex made it possible to generate, review, and refactor code faster than ever. But the process around them is still manual — prompts copied by hand, context lost between tasks, reviews and approvals tracked in your head. AutoPilot closes that gap, turning powerful AI coding tools into a managed software delivery workflow.

From AI coding chaos to controlled deliveryOne control center for tickets, reviews, approvals & logsYou stay in control at every step

AI agents on autopilot, with you in control.

Stop copy-pasting between Codex and Claude. Let AutoPilot run the loop.