By developers, for developers · Free and open source

Every repo. One window. Agents that finish sooner.

IntelyIDE is a fast, lightweight desktop IDE for people who work across several git repositories at once. It opens up to 7x faster than the most used IDEs on a basic system, and its agents start and finish up to 6x sooner. Agents can read and edit, but they never commit or push. You do.

Download for macOSVersion 1.10.0 · .dmg · 31 MB

Requires macOS 13.5 or later. One universal build runs on Intel and Apple Silicon Macs.

Other downloads and checksums
FileArchitectureSizeSHA-256
IntelyIDE_1.10.0_universal.dmg Universal (Intel and Apple Silicon) 31 MB b966c70eb790ff117447b315001922eee17802a6600d978589c25321eba8478b

All downloads and checksums

  • Multi-repo workspaceOne Changes tree for every repository.
  • Never commit or pushAgents edit. You commit and push.
  • Lead and workersA lead splits the task, workers share it.
  • Parallel runsSeveral runs at once, even on one repository.
  • MCP and DeepPatternYour tools and integrations in one catalog.

Results vary with workspace, machine and task.

IntelyIDE main window showing the multi-repository Changes tree next to an editor and an agent panel.

Open any workspace in moments

Many repositories, one window, and no long wait before you can work. IntelyIDE is a native shell built with Rust and Tauri, not Electron, and it shows your repositories as soon as they are read.

  • Native, not a browser in a boxA Rust and Tauri shell uses the system web view, so there is no bundled browser to start.
  • Ready with every repositoryThe window shows all repositories of the workspace together, in one Changes tree.
  • Light while you workDesigned to stay light and responsive on a basic Mac, not only on a fast one.
Time until a workspace is ready to useIllustrative
IntelyIDEup to 7x faster
The most used IDEsbaseline

Illustrative graphic of the “up to 7x” figure for startup with any workspace on a basic system. It is not a chart of one particular test. Results vary with workspace, machine and task.

Agents that start and finish sooner

The intelligent agent solution splits a task across agents, starts them together and keeps the expensive parts for the work that needs them. You see every run, and you decide what is committed.

  1. Lead and workers

    Start more than one run and Run 1 leads. It splits the task, hands each worker a subtask and gives you one combined summary. Only the lead asks you questions.

  2. Parallel runs

    Workers run side by side instead of one after another, so a task with independent parts finishes sooner.

  3. Runs open at once

    A new run opens right away, even when the Mac is busy, so you see it start instead of waiting in silence.

  4. Smart sub-agent settings

    Each role has its own model, effort and tools. The Auto lead delegates the simple parts to lighter roles, which is where the saving in tokens comes from.

  • up to 6xfaster agent start and finish thanks to the intelligent agent solution
  • about 30%fewer tokens thanks to the intelligent sub-agent settings

Results vary with workspace, machine and task.

Download IntelyIDE

Highlights

Why IntelyIDE

  • Many repos, one view

    Changes from every repository sit in one grouped tree. Review, commit and push across all of them without switching windows.

  • Agents that stay in their lane

    Agents can read and edit files. They cannot commit or push. Every run starts with a Rewind snapshot, so one click undoes it.

  • Light and fast

    A native shell built with Rust and Tauri, not Electron. Designed to stay light and responsive while you work.

What you get

Every repository’s changes in one tree

The Changes tree groups changes by repository, each with a repo badge, a branch pill and tri-state checkboxes. Pick what to stage, write one commit message for everything or one per repo, then push from a dialog with a checkbox for each repository.

  • Commit graph, blame and hunk staging
  • Interactive rebase, cherry-pick and stash
  • Pre-commit checks and a warning when a secret shows up in a diff
The Changes tree with four repositories and their changes grouped by repository.
The push dialog with one checkbox for each repository.

AI agents with guardrails

Agents run in roles, and each role has its own model and effort. They can read and edit files. They cannot commit or push. A “needs you” inbox collects every approval that is waiting for you.

  • Every run starts with a Rewind snapshot. One click undoes it.
  • Enforced in several layers: permission policy, hard stops, a git shim and hooks.
  • Uses your own Claude Code setup. The app guides you through it.

IntelyIDE does not ship Claude Code. Anthropic does not endorse IntelyIDE.

An agent safety card showing an approval waiting in the needs-you inbox.
The Rewind view showing the snapshot taken when an agent run started.

Workspaces and a welcome screen

A workspace is a named set of repositories. Keep as many as you like and switch between them. The welcome screen lists recent workspaces and lets you open a folder or scan a folder for repositories.

  • Workspace switcher
  • Open folder or scan for repositories
  • Worktrees and branch hygiene
The welcome screen with recent workspaces and the workspace switcher.

Run, preview, click to source

The run panel starts your dev server and shows a live preview next to the code. Click an element in the preview to jump to its source. Component preview is built in too.

  • Editor, terminal, search and diff in the same window
  • Live dev-server preview
  • Click-to-source and component preview
The run panel with a live dev-server preview and click-to-source.

Team: work together

Beta

Create a team on your own Cloudflare relay and invite colleagues with a one-time invite code. Both sides compare a six-digit safety code before anything is shared, and everything the relay carries is end-to-end encrypted. See what a teammate’s agent is doing while it happens, and help them without taking over.

  • Live view of a teammate’s agent runs: tool calls, answers and diffs, shared per project with secrets redacted
  • Chat in a run: a thread under every shared run, for the people looking at it. The agent never sees it.
  • Write into a teammate’s agent: a note while it works, a follow-up once it has stopped. It is marked as yours and reaches the agent only after its owner approves it.
  • A snapshot of a teammate’s workspace with their latest commits, and a team-wide commit list to see who is editing and committing what, opt-in per project
  • Shared Claude memory: memory/*.md notes stay in sync, CLAUDE.md is a separate opt-in with a review before any change
  • Teammates can never answer permission cards or start runs. Removing a member rotates the key.
The Team tab: a teammate's run read-only, with a note waiting for your approval.
Chat in a teammate's shared run: the thread belongs to the people, the agent never sees it.
The commits and edits of the whole team, with a timeline of everybody's recent commits.
Shared Claude memory with a pending CLAUDE.md change and a conflict copy.

Run agents on your own servers

Tried against a Linux container and a scripted stand-in, not yet many real servers

Put a big machine to work. Add a server by its ssh target and the IDE sets it up for you: Node.js if it is missing, the agent files and the Agent SDK. Then start runs there from the same New run dialog.

  • Split a task, for example 3 agents on the build server and 2 on this Mac
  • A limit of agents per server
  • The permission rules and the approval broker stay in the IDE and judge the server’s files
  • The git guard travels with each run, so agents there cannot commit or push either

Phone gateway: approve from anywhere

Experimental: not yet verified against a real Cloudflare account or iPhone

Pair your phone with the IDE through a relay you deploy on your own Cloudflare account. Watch agent runs, and approve or answer them when you are away from your desk. Everything is end-to-end encrypted, and the project hosts no relay for you.

  • Pair a phone with a one-time code and a safety code you confirm on the Mac
  • Approve a permission or answer a question from the phone
  • Optional and off until you set it up
The phone companion showing an agent run in progress.

Notifications: know when a run needs you

Start a long run, switch to something else and let the IDE tap you on the shoulder. A banner appears when a run needs a permission, asks a question, finishes or fails, and the Dock icon counts the runs that are waiting.

  • A switch for each kind of notification
  • A gap between banners and a cap per minute, so you are never flooded
  • An optional sound

Integrations: connect your tools in one place

Settings > Integrations is a catalog of the services the IDE connects to. Happy, Sentry and DeepPattern show whether they are connected, nothing connects until you set it up, and secrets stay in the macOS Keychain or with the installer of the service.

  • Happy: chat, tasks, Time Tracer, Meet and notifications of your Happy account, next to your code. Choose per kind (orders, delivery, invoices and more) what alerts you
  • MCP servers: more tools for agents, over stdio or HTTP, with a rule for every tool
  • Sentry: read your issues and hand one to an agent to fix
  • DeepPattern: second opinions on your code and plans from a panel of models. Install and activate it with its own installer (the key never enters the IDE), check it, add it to your agents and audit your changes. A new service is one more card in the same catalog
Settings, Integrations: cards for Happy, MCP servers, Sentry and DeepPattern; each card shows its connection status.

MongoDB Studio

Experimental: not yet verified against every real MongoDB setup, such as Atlas

An optional MongoDB browser that is read-only by default. Describe the query you want, in English or in Hungarian, and the AI drafts it for you to review.

  • Read-only by default
  • AI-assisted query generation, including Hungarian prompts
  • SSH tunnels and TLS
MongoDB Studio showing a read-only query and its results.

API contract checks

Backend and clients drift apart quietly. The API contract extra compares your backend’s Swagger with the client repositories and flags where they no longer match. A read-only Swagger explorer lets you browse the spec without leaving the IDE.

  • Contract-drift findings across repositories
  • Read-only Swagger explorer
  • Optional: switched on in Settings. While it is off, nothing is loaded or scanned.
The API contract view with the Swagger explorer and contract-drift findings.

Works with

IntelyIDE connects to the services you already use and shows where each one stands, next to your code. Two of them have their own card in Settings > Integrations.

  • DeepPattern.ai

    Second opinions on your code, plans and ideas from a panel of models from several vendors. Install the Decision Engine with its own installer, activate it with your key (the key never enters the IDE), check it, add it to your agents and audit your uncommitted changes. DeepPattern is available by invitation.

    Visit DeepPattern.ai

  • Happy Solutions

    Chat, tasks, Time Tracer, Meet and the notifications of your Happy account, next to your code. Choose per kind of notification (orders, delivery, invoices and more) whether it alerts you, stays quiet or is off.

    Visit Happy Solutions

Names and logos belong to their owners and are shown only to name the services you can connect. No endorsement is implied.

Bring your own agent

Claude is the main path. Other agent tools plug in as experimental providers that stay off until you switch them on.

  • ClaudeMain path
  • CodexExperimental
  • GeminiExperimental
  • GitHub CopilotExperimental
  • OpenCodeExperimental
  • GooseExperimental
  • QwenExperimental
  • Custom ACP agentExperimental
  • One global “Experimental providers” switch, off by default
  • You confirm each provider’s command line before it can run
  • Non-Claude providers are read-only by default and show an honest safety rating
  • Your own installs and accounts. IntelyIDE ships none of these tools.

Experimental: the experimental providers have not yet been verified against the real programs.

Settings, Providers: the provider cards and the Experimental providers switch.

And the rest of the toolbox

  • Editor and terminal
  • Search and diff
  • Pre-commit checks
  • Secret-in-diff warning
  • Branch hygiene
  • Worktrees
  • Localization checker
  • Release assistant
  • Session search
  • Night queue and morning brief
  • 53 interface languages
  • 11 accent colors and a custom one
  • Light and dark themes

English and Hungarian are hand-written. The other 51 languages are machine translated.

Built for trust

An agent can propose and edit. Only you commit and push.

  1. AgentReads and edits files in your repositories.
  2. PolicyPermission policy, hard stops, a git shim and hooks check each action.
  3. Your approvalAnything that needs a decision lands in the needs-you inbox.
  4. You commit and pushReview the diff, then commit and push yourself.

One click back

Every agent run starts with a Rewind snapshot. If a run goes the wrong way, restore the snapshot.

Best effort, not a guarantee

These layers work together, but they are best effort. They are not a security boundary guarantee. Read what an agent changed before you commit it.

Local-first

Your work stays on your machine. Credentials are stored in the macOS Keychain. IntelyIDE has no telemetry. Once a day, after a one-time notice, it asks the public GitHub API whether a newer release exists and sends nothing but its version number. You can turn that off.

Signed and notarized

The app and the disk image are signed with a Developer ID and notarized by Apple, so macOS opens IntelyIDE on the first launch without a warning, even offline.

Free and open source

IntelyIDE is licensed under GPL-3.0-or-later and free to download. Read the code, build it from source, report a bug or send a pull request.

Contribute

Pull requests are welcome. The maintainer reviews and approves every pull request manually before it is merged.

Install in 3 steps

  1. Download the .dmg

    Use the Download for macOS button. Optionally check the SHA-256 checksum first.

  2. Drag to Applications

    Open the .dmg, then drag IntelyIDE onto the Applications folder.

  3. Open it

    Start IntelyIDE from Applications. It will guide you through connecting your own Claude Code setup if you want agents.

Signed with a Developer ID and notarized by Apple. It opens without warnings.

Download for macOS

Questions

Is it free?

Yes. IntelyIDE is free to download and open source under GPL-3.0-or-later.

Does it send my code anywhere?

IntelyIDE itself has no servers. The agent features use your own Claude Code setup, which talks to Anthropic under your own account and terms. Remote is optional, end-to-end encrypted, and runs through a relay on your own Cloudflare account.

Does it work without AI?

Yes. Agents are optional. The multi-repo Changes tree, editor, terminal, search, diff and the rest work without one.

What do I need for the agents?

Your own Claude Code setup. The app guides you through connecting it. Other agent tools, such as Codex or Gemini, can be added as experimental providers. IntelyIDE ships none of them.

Does it run on Apple Silicon?

Yes. The current build is universal, so it runs natively on Intel and Apple Silicon Macs. The architecture note next to the download button always shows what the current release is.

What about Windows and Linux?

Not right now. IntelyIDE is built and tested on macOS only.

How do I report a problem?

Open an issue on GitHub. Say which version you use and what you did before it happened. Questions go to GitHub Discussions. Report security problems privately through GitHub’s private vulnerability reporting.

Is it affiliated with Anthropic?

No. IntelyIDE is an independent project and is not affiliated with or endorsed by Anthropic.

More from IntelyHome

IntelyOpti

A native optimizer for Intel Macs: cleanup, Turbo, network speed, Mac Health and an AI Advisor. 14-day free trial.

Explore IntelyOpti