Let AI move fast.
Keep your code safe.
AgentHub records every AI turn, hands off context between agents, and restores your workspace in one click — before AI mistakes break your code.
Windows 10/11 · Native performance · No telemetry · Cancel anytime
Full product walkthrough · 90 seconds
/demos/demo-all.mp4
Supports the IDEs you already use
Cursor
Electron
Antigravity
Claude Code
MSIX Store
Codex
OpenAI
VS Code
+ Copilot
Windsurf
Codeium
Plus any custom .exe · MSIX Store apps · Win32 native
The problem
Three things that should not still be hard in 2026.
Modern AI-assisted development means juggling 3-5 IDE processes daily. Every one of them is a separate window. None of them remember what the others did. Nothing about your workflow has a real undo button.
Window chaos
Five Cursor windows. Two Claude Code instances. VS Code somewhere in the back. Your taskbar is a graveyard, every Alt+Tab is a roulette spin.
Fixed by → Hub
Memory amnesia
Cursor refactored auth. Claude Code has no idea any of that happened. You have no idea what AI did 47 minutes ago. Re-explaining context is your full-time job.
Fixed by → Continuity + Timeline
No real undo
AI rewrote 12 files in 30 minutes. Half compile, half don't. Git status is a horror movie. There's no clean rollback to that one moment things were fine.
Fixed by → Turn restore + Workspace snapshot
Five things AgentHub does for you
The safety net for AI-assisted coding.
Every AI IDE in one window — that's table stakes. The real value: full memory across agents, every turn searchable, every change reversible.
/demos/feature-hub.mp4
Feature 01
Every AI IDE. In one tabbed window.
Five Cursor windows. Two Claude Code instances. Where did your VS Code go? Your taskbar is a graveyard, your Alt+Tab cycle is unbearable.
AgentHub embeds Cursor, Antigravity, Claude Code, Codex, VS Code, and Windsurf into a single tabbed Windows app. Switch tabs in 150 ms with a zero-flash crossfade — no flicker, no white flash, no relogin. Each IDE renders natively, so performance stays exactly as you remember it.
- 150 ms zero-flash tab switch with snapshot crossfade
- Native Win32 reparenting — no rendering interception
- Auto-detects every supported IDE on first launch
- MSIX Store apps via COM activation (no UAC prompts)
- Single Alt+Tab entry — sub-IDEs hidden via WS_EX_TOOLWINDOW
/demos/feature-continuity.mp4
Feature 02
Your AI agents understand each other.
You spent 20 minutes in Cursor refactoring auth. You switch to Claude Code — it has zero memory of any of that. Re-explain everything. Again.
AgentHub captures every AI conversation across IDEs and writes a structured handoff log. When you switch from Cursor to Claude Code, the next agent reads exactly what the previous one did — files changed, decisions made, current state. No more re-explaining context to each new agent.
- Auto-parses transcripts from Cursor, Claude Code, Codex, Antigravity, Copilot
- Writes a structured continuity log to .agenthub/ on every IDE switch
- Includes file changes, tool calls, and conversation summaries
- MCP server provisions per workspace — agents read context via tool calls
- Mark blocks as verified to keep them; old ones auto-prune
/demos/feature-timeline.mp4
Feature 03
Every AI turn. Indexed. Searchable.
Cursor wrote that auth module 47 minutes ago. Or was it Claude Code two hours ago? You don't remember. Your terminal is a wall of text. Scrolling won't help.
AgentHub records every AI turn — prompt, response, tool calls, files touched — and surfaces them in a unified timeline panel. Filter by IDE, search by content, color-coded per agent. Click any turn to inspect the exact tool calls and the precise code that changed.
- Side panel timeline shows turns from every embedded IDE
- Color-coded per IDE (Cursor purple, Claude red, Codex teal…)
- Search across prompts, responses, tool calls
- Highlights show exactly which files and regions each turn touched
- Doesn't break your IDE flow — collapsible at any moment
/demos/feature-restore.mp4
Feature 04
Roll back any AI turn. In one click.
AI made the right change 30 turns ago. The next 29 turns built on top of it incorrectly. Your codebase is layered with mistakes. There is no clean undo.
Every AI turn captures pre-state and post-state of every file it touched. Open any turn → review the exact diff → click Restore to revert your codebase to that moment. Like git, but automatic and turn-aware. No manual commits required.
- Pre/post state captured automatically for every turn
- Side-by-side diff viewer for each turn's changes
- Restore individual files or roll back to a turn-level state
- Inspect every tool call the agent made
- Content-addressed storage — efficient even with thousands of turns
/demos/feature-snapshot.mp4
Feature 05
Snapshot before AI breaks things.
You let Cursor refactor for 30 minutes while you got coffee. Came back to half the codebase rewritten, nothing compiles, and no clean rollback. Now you panic.
AgentHub auto-snapshots your workspace at intervals you choose, and lets you take manual snapshots before risky AI prompts. When AI goes sideways, restore the entire workspace to a known-good state in seconds. A safety net you can actually trust.
- Auto-snapshot at configurable intervals (e.g., every 10 min)
- Manual snapshot before any prompt you don't fully trust
- Full workspace restore — every file, every state, instant
- Storage cap with smart LRU retention so disk stays bounded
- Atomic writes — snapshots never corrupt mid-save
How it works
Three steps. No configuration.
Install once, point at a workspace, work like you always do — except every IDE you open shares context with the next.
Open a workspace
Pick a project folder from the welcome screen — or jump back to a recent one. AgentHub remembers what you were working on.
Pick an IDE
Hub auto-detects every IDE you have installed — Cursor, Antigravity, Claude Code, Codex, VS Code, Windsurf — with icons and versions ready.
Work like normal
Switch between tabs instantly. Context and chat history sync across agents through .agenthub/ logs. The next agent always knows what the previous one did.
Pricing
Simple, transparent pricing.
One subscription unlocks every supported IDE on every device you own. Try free for 7 days — no credit card required.
Monthly
Flexible · pay as you go
- Every supported AI IDE
- 1 device per license · move anytime
- Cross-IDE conversation continuity
- AI turn timeline + workspace snapshots
- Auto-update via WinGet
- Email support
Cancel anytime · No card needed for trial
Yearly
Save 25%
Billed $71.88 once a year
- Everything in Monthly
- 1 device per license · move anytime
- Priority email support
- Early access to new features
- Roadmap voting
$71.88 charged once a year · save $24 vs monthly
Lifetime
Pay once · never again
- Everything in Yearly
- 1 device per license · move anytime
- All future updates included
- Founder badge in app
- Zero recurring billing — ever
Limited early-bird pricing · ~22 months payback vs monthly
14-day money-back guarantee · Pricing in USD · VAT/tax handled by Polar
Questions
Frequently asked
Can't find what you're looking for? Email support@ideagenthub.com — we reply within 24h.
Windows 10 (build 1809+) and Windows 11. AgentHub uses native Win32 APIs and DirectComposition. macOS and Linux are not supported and there are no plans to support them — the technical foundation is Windows-specific.
No. AgentHub uses documented Win32 APIs (SetParent, SetWindowLongPtr, Job Objects) to host the IDE's main window inside our container. The IDE renders natively — we never intercept its rendering pipeline. Detaching restores the IDE to its original state.
Yes. Every IDE login continues to work normally because the IDE process runs natively. AgentHub doesn't touch your account credentials. If Cursor is already running when you launch AgentHub, we attach to your existing session instead of spawning a duplicate.
AgentHub watches each IDE's transcript files (Cursor JSONL, Claude Code session files, Codex events, Antigravity Language Server, Copilot details). Every AI turn is parsed into a structured record with prompt, response, tool calls, and the files it touched. Records appear in the side panel timeline, color-coded per IDE. Storage is content-addressed so it stays compact.
Yes. AgentHub takes auto-snapshots at intervals you set (default off — you opt in) and you can take manual snapshots at any time. A snapshot captures every file in your workspace (with sensible ignore rules for node_modules, .git, build outputs). Restoring writes everything back atomically. There's a configurable storage cap with LRU retention so disk stays bounded.
When you switch IDEs, AgentHub appends a structured handoff block to .agenthub/continuity-log.md in your workspace — what the previous IDE just did, files changed, key decisions. Per-workspace MCP server provisioning means agents read this context via tool calls, not raw file scraping. Old blocks auto-prune after they're marked verified.
Yes. Add any executable with a display name through Settings → Custom Tools. Useful for in-house agents, custom CLI wrappers, or IDEs not yet supported out of the box.
Each license is bound to your hardware fingerprint (HWID = SHA256 of MachineGuid + Motherboard UUID + CPU ID). Every plan — Monthly, Yearly, Lifetime — includes 1 active device at a time. You can revoke the active device from your account dashboard and activate a new one in seconds, so moving between work and home machines is easy. Online validation runs every 24 hours; 14-day offline grace period if you're traveling.
Your access continues until the current billing period ends. After that, AgentHub switches to read-only mode — you can still view existing tabs, the turn timeline, and snapshots, but new IDE tabs cannot be spawned. Resubscribe anytime to restore full functionality.
No telemetry from the desktop app. We track aggregate purchase events through Polar (our payment processor) and basic activation events on our license server. Your code, prompts, AI conversations, and IDE state never leave your machine.
14-day money-back guarantee, no questions asked. Refunds processed by Polar (Merchant of Record). Email support@ideagenthub.com to request.
Move fast. Without breaking things.
Try AgentHub free for 7 days. No credit card. Cancel with one click. Keep every AI IDE you already use — and the safety net you didn't know you needed.
Windows 10/11 · 200 MB · Install via WinGet or direct download