If you write code for a living, you've probably tried a bunch of AI coding tools by now. Claude Code, Codex, OpenCode, Grok Build — they each have their strengths. But until recently, using them meant picking one and committing to it, or juggling multiple terminal sessions like it was 1999.
That's the problem Paseo solves. It's an open-source desktop and mobile app that acts as a control plane for every coding agent out there. You run them all — simultaneously — in one interface.
Multiple agents, multiple tabs, one window
Paseo supports Claude Code, Grok, OpenCode, Codex, Cursor, Pi, and about 34 other providers. You can open a Claude Code tab to architect a feature, a Codex tab to implement a component, and an OpenCode tab for code review — all at the same time.
This is the killer feature. When you hit a wall with one agent's reasoning style, you switch to another in the same project context without losing your flow. The agents share the same working directory, so they can build on each other's work. I've started a refactor in Claude Code, had Codex handle the boilerplate, and used Grok to write the tests — all in parallel tabs.
Built-in terminal that actually works
Every project in Paseo gets a built-in terminal alongside your agent tabs. It's not a half-baked embedded shell — it's a real terminal with everything you'd expect. Run your dev server, check logs, hit APIs, install packages, all without leaving the app. The agents can see the output too, which makes for much tighter feedback loops.
Code on the go — really
The mobile app is what sold me. You get a full desktop app (Linux, macOS, Windows), and it pairs with your phone. I've started a coding session on my laptop, stepped out, and continued the exact same agent conversation on my phone via the mobile app. The project context, the conversation history, the terminal output — it all syncs.
This sounds like a gimmick until you actually need it. Waiting for a build to finish? Pull out your phone and review a diff. On a train with your laptop put away? Open the web app and check on that long-running agent task. It's the same Paseo session, just on a different screen.
Open source, self-hosted, free
Paseo is MIT-licensed and free. No paid tiers, no usage limits, no data leaving your machine unless you want it to. It's self-hosted by default — the desktop app runs everything locally. If you want remote agents (run code on a beefy cloud box while you work from a thin laptop), that's built in too.
There's a plugin system, an MCP interface, a CLI, and an SDK. It's designed to be extended. If Paseo doesn't support something you need, you can build it.
Why it's my go-to
I've tried pretty much every coding agent setup you can think of. Claude Desktop for quick tasks, the CLI for deeper work, Cursor as an IDE, custom MCP setups, chaining scripts together. Paseo replaces most of that complexity with a single app that does the mixing and matching for me.
The multi-agent tab model is genuinely better than anything else I've used. Different agents have different strengths — Claude Code is great at reasoning and planning, Codex is fast at implementation, Grok Build handles tool-heavy workflows well. Paseo lets me pick the right agent for the moment without context-switching overhead.
Head over to paseo.sh. Download the desktop app, grab the mobile app, or use the web version. It's free, open source, and it'll change how you think about coding agents.
GitHub repo at github.com/paseo-io/paseo — 15.7k stars and growing. Give it a shot.
No comments yet