AI Labs
Your computer, from anywhere.
AI Labs (cptr, formerly HridaAI Computer) is a standalone app that runs on your machine and serves your whole computer to any browser: files, terminal, editor, git, browser tabs, running sessions, AI agents, and tools. It literally is your computer, reachable from your phone, tablet, laptop, or any other device.
Because sessions, files, and agents keep running on your machine, AI Labs doubles as a persistent agentic runtime: R&D and engineering teams use it for long-horizon research and coding work that keeps going after you close the browser.
Bring your own API key, plug in a coding agent subscription you already have, or drive it through the Gateway API from Hrida.ai or another OpenAI-compatible client. One tool, full workstation, any device.
AI Labs is a different product from Hrida Terminal. Hrida Terminal is the tool-execution backend built into Hrida.ai chat — it gives an existing Hrida.ai conversation a sandboxed environment to run code in. AI Labs is a separate, standalone workstation app you run on your own machine and open directly in a browser; it can also expose itself as an OpenAI-compatible model that Hrida.ai (or any other client) connects to via its Gateway API.
Why it exists
The repo you already cloned. The branch you forgot was still dirty. The terminal that is still running. The .env file you made three months ago and never want to recreate. That is the computer you trust — and the moment you step away from the desk, most tools ask you to leave it behind.
AI Labs' premise is the opposite: you should be able to reach your computer without moving your work off your computer. A build finishes while you are making coffee. A deploy fails while you are on the couch. An agent finishes a task while you are walking outside. Look at the output, read the diff, type one sentence, press the button — from the screen in your pocket. The full read of the philosophy is in the project's MANIFESTO.md.
Use cases
Some of the things people actually reach for AI Labs to do:
-
Let it tidy a folder you've been putting off. Point the agent at a Downloads folder that's turned into a junk drawer and have it sort, rename, and dedupe the mess into something you can actually navigate. See The workstation.
-
Open the same project no matter which screen found it. Workspaces, open tabs, and terminal state aren't tied to one browser — start on the desktop, continue from a tablet, finish on your phone. See The workstation.
-
Build a first website with the agent looking over your shoulder. Write a page, watch it render in the built-in browser next to the editor, and iterate out loud instead of guessing what the markup will look like. See AI agent.
-
Text your computer instead of SSHing into it. Ask your Telegram bot to check a log or restart a service, and get the answer back in the same chat thread, no terminal in sight. See Messaging bots.
-
Push a one-line fix before you're back at your desk. Open the file, make the edit, commit, and push — entirely from the phone in your pocket, no laptop required. See The workstation.
-
Catch a layout bug on an actual phone, not a simulator. Point the in-workspace browser at your local dev server from your own mobile device and see exactly what a real visitor would see. See The workstation.
-
Get pinged before you think to go looking. Wire a messaging bot to a webhook or a scheduled check so it messages you the moment something needs attention, instead of you polling for it. See Messaging bots.
-
Walk away from a multi-step job, come back to a finished result. Hand a sub-agent the whole task — research, draft, revise — and check back once, when it's actually done. See Sub-agents.
-
Stop re-explaining the same context every session. Once it's told which linter to skip or how your team names branches, it recalls that on its own the next time it's relevant. See Memory.
-
Watch a coding agent work without sitting at the desk it's running on. Kick off Claude Code or Codex against a real task and follow its reasoning and diffs from wherever you are. See Coding agents.
-
Start the day with the status report already written. Schedule a recurring job that checks overnight activity and has a summary waiting before you sit down. See Automations & scheduled tasks.
-
Run parallel agents on git worktrees. Check out the same repo into separate worktrees and point a different agent at each branch, so three fixes move at once without their contexts colliding. See Sub-agents.
The whole machine
Files, shell, git state, workspaces, chats, tools, and sessions stay together wherever you open it.
| 📁 Real files | Navigate, create, rename, upload, drag and drop. |
| ✏️ Editor | Syntax-highlighted editing with tabs, side by side. |
| 🔀 Git | Stage, commit, diff, branch, push — no command line required. |
| ⌨️ Terminal | Full shell in the browser. Run your tools, scripts, or a coding agent. |
| 🌐 Browser | Proxy, a private Managed Chrome profile, or your own Personal Chrome session. |
| 🔄 Sessions persist | The terminal keeps running when you close the tab. |
| 🗂️ Tabs | Browsers, terminals, files, chats, and tools in separate, rearrangeable tabs. |
| 📂 Workspaces | Multiple projects, one instance. Switch without losing your place. |
| 🔍 Search | Find files by name or contents; jump straight to a matching line. |
| 📱 Mobile-first | Built for the screen in your pocket, not a shrunk-down desktop UI. |
See The workstation for the full surface.
Feature map
| Area | What's there |
|---|---|
| The workstation | Files, editor, git, terminal, in-workspace browser, tabs and layout, workspaces, search, inline previews, document reading, PWA. |
| AI agent | Streaming chat with tool calling, reasoning display, plan mode, output editing, @/$ mentions, context compaction, voice mode, text-to-speech, voice memos, message queue, image understanding and generation, per-response usage. |
| Tools & tool servers | Built-in file / terminal / web / browser / memory / automation / image tools, external tools over MCP or OpenAPI, tool approval and Auto mode. |
| Coding agents | Codex, Claude Code, Cursor, Grok, OpenCode, Cline, Gemini, Pi as native backends using subscriptions you already pay for. |
| Sub-agents | Parallel background workers with full tool access. |
| Skills | Reusable SKILL.md instruction sets, $-mentioned, created from chat, kept up to date automatically. |
| Memory | File-backed managed memory with automatic recall and background review. |
| Automations & scheduled tasks | Recurring AI jobs on an RRULE schedule, webhook-triggered runs. |
| Messaging bots | Telegram, Slack, WhatsApp, Signal with full tool access. |
| Gateway API | Each workspace as an OpenAI-compatible agent model at /v1/chat/completions. |
| Observability | Tamper-evident activity log, audit logging, diagnostic logging, notifications, health checks. |
| Administration | Admin settings, RBAC roles, system prompts, 10 languages. |
| Authentication & access | Username / password, JWT sessions, trusted-header sign-in, signup toggle, API keys. |
| Installation · Docker · Configuration | Install, run, deploy, air-gap, and tune. |
| Security model | What "full machine access" means and when it is safe. |
Install
pip install cptr
cptr runOpens in your browser at http://localhost:8000. Full options — uv, phone access, tunnels — are in Installation; container and air-gapped deployment are in Docker.
License
Open Use License. Source available. Commercial licenses and enterprise licenses are available. See Security model for details.