Your ChatGPT conversation and requests to Rel.AI.
Let ChatGPTwork with yourlocal code.
Rel.AI connects ChatGPT web to your local project so it can find files, edit code, run commands and tests, review changes, and use Git.
work_7f3a
@media (max-width: 680px) {
- .nav-links { gap: 14px; }
+ .nav-links { grid-template-columns: 1fr; }
+ .nav-links a { min-height: 44px; }
}
Desktop render1440 × 900
passMobile render390 × 844
passConsoleno runtime errors
passYour project stays local. Files, commands, running processes, and Git actions stay on the computer you selected.
Each task stays separate. Every new goal gets its own work session, so its changes and checks stay easy to follow.
You can review the result. Rel.AI checks the work, and it never publishes changes unless you ask.
From your ChatGPT request to a checked code change.
Rel.AI gives ChatGPT a clear coding workflow: understand your project, make focused edits, run it, check the result, and publish only when you ask.
- 01Understand
Find the right files and learn how they connect before reading or changing code.
snapshot · search · inspect · read - 02Change
Make focused edits inside the selected project without overwriting newer changes.
edit - 03Run
Run commands, start local servers, and interact with the app when needed.
exec · process · ui - 04Check
Run tests and inspect the app to confirm the latest changes work.
validate · changes - 05Publish
Commit or push changes, or prepare pull-request text, only after review and only when you ask.
publish
Keep using ChatGPT. Let it work directly with your local project.
Rel.AI is for developers who already like working in ChatGPT. Instead of copying code back and forth, ChatGPT can read your project, make edits, run it, and check the result on your computer.
See how Rel.AI worksFind the relevant files yourself.
Paste snippets into the conversation.
Apply suggestions in another tool.
Run and report checks manually.
ChatGPT asks Rel.AI to look at the project you selected.
Rel.AI finds and reads only the files needed for the task.
ChatGPT makes focused edits and runs the project on your computer.
Rel.AI checks the latest changes and publishes only when you ask.
A small set of tools for everyday coding work.
ChatGPT gets 12 focused tools for reading code, editing, running your app, checking changes, and using Git. Rel.AI handles permissions, task tracking, and activity history in the background.
See the coding workflowFind and read the right code before changing it.
relai_workrelai_snapshotrelai_readrelai_searchrelai_inspectEdit the project and run it on your computer.
relai_editrelai_execrelai_processrelai_uiConfirm the result, then use Git only when you ask.
relai_validaterelai_changesrelai_publishChatGPT sends the request. Rel.AI runs it on your computer.
The OpenAI Secure MCP Tunnel connects ChatGPT to Rel.AI privately. Rel.AI then works only in the local projects you added and keeps track of changes, checks, running processes, and Git actions.
Runs approved tools on your computer, keeps each task separate, limits work to your project, and records checks.
ready on this computerFiles, commands, tests, running apps, browser sessions, code changes, and Git.
ChatGPT can work only inside folders you add as Rel.AI workspaces.
Changes, checks, review, recovery, cancellation, and completion stay linked to the same task.
A tool saying “done” is not enough. Rel.AI checks the current code before the task is considered complete.
Editing code does not automatically commit, push, or create pull-request text.
ChatGPT and the ChatGPT logo are trademarks of OpenAI. Logo used only to identify the supported ChatGPT integration.
Code in ChatGPT. See local activity in the Rel.AI desktop app.
The Rel.AI desktop app runs the local service and shows your workspaces, active tasks, tool activity, running processes, connection status, usage, updates, and diagnostics.
Set up the desktop appwebsite
Project searched8 relevant matches · 10:14:04
relai_searchResponsive styles edited2 files changed · 10:14:17
relai_editMobile UI session390 × 844 · running
relai_uiYour code and commands stay on your computer.
Your project files and commands stay on the computer you selected. The Secure MCP Tunnel only connects ChatGPT to Rel.AI; Rel.AI keeps your workspace limits, credentials, tasks, processes, checks, and Git actions local.
Technical security detailsRel.AI can run code in projects you add. Only add projects you trust ChatGPT and Rel.AI to read and change.
Connect ChatGPT to your first local workspace.
Set up the desktop app once, connect it to ChatGPT with the OpenAI Secure MCP Tunnel, then choose which local projects ChatGPT can work on.
- 01Install Rel.AI
Download the desktop build for Windows, macOS, or Linux.
Open releases - 02Start the secure connection
Enter your Tunnel ID and API key. Rel.AI uses them to run the included OpenAI tunnel client.
See connection flow - 03Add a workspace
Choose a local project folder and give it a short name such as
Workspace safetymyapp. - 04Use Rel.AI in ChatGPT
Turn on Rel.AI MCP in ChatGPT, then describe what you want changed in normal language.
Try a prompt
Use Rel.AI MCP with workspace "myapp". Start one work session, read the project, and explain how the relevant parts work before changing anything.
Rel.AI is open source.
You can review the code that connects ChatGPT to your projects, handles permissions, runs local tools, checks changes, and uses Git.
What developers usually want to know.
Rel.AI gives normal ChatGPT conversations a Codex-style local coding workflow without using the Codex agentic allowance. Here is what that means in practice.
Can Rel.AI replace Codex?
Yes, for the workflow Rel.AI is built for. If you use Codex mainly to inspect a repository, edit code, run commands, validate changes, and use Git, Rel.AI gives ChatGPT those capabilities on your own computer. It does not emulate Codex or use Codex internally; it gives normal ChatGPT conversations a Codex-style local coding workflow.
Does Rel.AI use my Codex limit?
No. Rel.AI runs through ChatGPT's app and tool path. OpenAI currently counts Codex toward agentic usage, while ChatGPT Apps follow the normal ChatGPT limits for your plan. Rel.AI therefore does not draw from your Codex agentic allowance. Your normal ChatGPT plan limits still apply.
What model does Rel.AI use?
Rel.AI does not provide a separate AI model. ChatGPT remains the reasoning layer, using the app-compatible model and reasoning mode available on your plan. On eligible plans that includes GPT-5.6 Sol reasoning modes such as High and, where available, Extra High. ChatGPT Apps do not support Pro models, so GPT-5.6 Sol Pro is not available through Rel.AI.
Do I need to copy and paste files into ChatGPT?
No. Rel.AI can search the workspace, inspect project structure, follow code relationships, and read only the files or line ranges needed for the task. That lets ChatGPT gather context from the repository instead of making you paste code back and forth.
Does my code stay on my computer?
Your repository, commands, Git operations, tests, builds, running processes, workspace settings, and Rel.AI activity history remain on the computer running Rel.AI. ChatGPT receives the bounded file content, search results, diffs, command output, and other tool results it needs through the Secure MCP Tunnel. Rel.AI does not move your repository into a hosted coding VM.
Can ChatGPT access anything on my computer?
No. Rel.AI only works inside folders you explicitly add as workspaces. It blocks path traversal, arbitrary absolute-path access, and symlink escapes, and it applies stricter handling to sensitive paths and secret-like content.
Can ChatGPT run my app and check the result?
Yes. Rel.AI can run project commands and tests, keep development servers and watchers running, open supported browser sessions, interact with the UI, inspect console and network activity, take screenshots, and verify the result. The execution happens on your computer.
Will Rel.AI commit or push code without asking?
No. Editing and publishing are separate. Rel.AI can review changes, commit scoped work, push to an existing remote, or prepare pull-request text, but those publishing actions stay explicit instead of happening automatically after an edit.
Why is Rel.AI built specifically for ChatGPT?
Because Rel.AI is meant to make the normal ChatGPT conversation you already use practical for real coding work, without requiring Codex for every repository task. OpenAI currently documents that ChatGPT Apps use the normal ChatGPT rate limits for your plan, while Codex counts toward agentic usage. Rel.AI therefore gives a regular ChatGPT conversation local repository tools without drawing from your Codex agentic allowance. Your normal ChatGPT limits still apply, so we do not describe it as unlimited. The advantage is that your general ChatGPT usage and your Codex allowance are not the same coding meter. That differs from products such as Claude Code, where Anthropic documents that Claude and Claude Code share the same Pro or Max usage limits.
Does Rel.AI support Claude, Cursor, Gemini, or other AI clients?
Not currently, and that focus is intentional. Rel.AI supports ChatGPT through the OpenAI Secure MCP Tunnel instead of trying to be a generic bridge for every AI client. Keeping one supported integration lets the desktop app, work-session model, workspace permissions, validation, recovery, and publishing rules be designed and tested as one complete workflow. Other AI products also use their own plan limits and quota systems, so supporting them would not automatically provide the same usage model that motivated Rel.AI.