Paseo is moving like a product that found a sharp pain: coding agents are useful, but babysitting them from one machine is dumb. ToolVitals saw 22 release events in 30 days and 30 GitHub releases in 90 days, with a shipping score of 100 and 5,542 GitHub stars. That is not quiet maintenance. It is an early tool hardening the messy parts of remote agent work.

The official site frames Paseo as a free, open-source, self-hosted way to run Claude Code, Codex, OpenCode, Copilot, and Pi from desktop, web, mobile, or CLI. The GitHub README uses the same core pitch: agents run on your own machine, while clients connect from phone, desk, browser, or terminal.

The signal is not just release volume

The interesting part is what shipped. Paseo v0.1.60 added scripts and services per worktree, setup progress, GitHub checks and PR review visibility, per-remote-host provider configuration, direct Pi integration, and beta channels. That points at a specific bet: agent orchestration is not only chat. It is process management, branches, diffs, PR state, ports, setup hooks, and recovery when something weird happens.

The worktrees docs make that bet explicit. Each agent runs in its own Git worktree so parallel agents do not step on each other. A paseo.json file can define setup, teardown, scripts, and long-running services. Services can get ports and deterministic localhost URLs, which matters when multiple agents are each trying to run npm run dev without turning your machine into port soup.

The later releases look like polish on that same bet. v0.1.61 added provider docs and an additionalModels option for extending model lists. v0.1.62 added app and daemon version drift warnings and fixed packaged CLI installation from the macOS desktop app. v0.1.63-beta.1 focused on instant worktree archiving, steadier terminal recovery, calmer timelines, and line-number file links from assistant messages.

That sequence is a good sign. The team is not only adding headline features. It is sanding down the operational rough edges that appear when agents are long-running, remote-controlled, and tied to real repos.

What ToolVitals can and cannot infer

ToolVitals can say Paseo is shipping quickly. The scorecard shows a 94 ToolVitals score, 92 health score, 100 shipping score, 216.2 hot score, 5,542 GitHub stars, 22 release events in 30 days, and 30 GitHub releases in 90 days. The data confidence is 82, so the signal is useful but not perfect.

ToolVitals cannot say Paseo is reliable in your repo, that its mobile app feels good every day, that its agent handoffs produce better code, or that teams are adopting it in production. Stars and releases are activity signals. They are not user satisfaction, revenue, code quality, or proof that the product works well under your exact workflow.

There is also missing data. The payload does not include 30-day commit counts or active contributor counts. That means release cadence is the strongest visible activity signal here, not a full picture of development breadth.

How Paseo compares

Paseo is much smaller than LangChain by stars, 5,542 versus 135,943, and it shipped fewer 30-day release events, 22 versus 38. That comparison is not a knock. LangChain is a broad framework brand. Paseo is narrower: remote orchestration for coding agents across devices.

Gemini CLI is closer in category momentum, with 26 release events in 30 days and 103,265 stars. Paseo trails it on audience size, but the product shape is different. Gemini CLI is an agent surface. Paseo is trying to become the control plane around multiple agent surfaces.

ToolJet is a useful counterpoint. It has 37,868 stars and 18 release events in 30 days. Paseo has fewer stars but more recent release events, which suggests a smaller project with unusually concentrated shipping tempo.

Recommendation

If your team already runs Claude Code, Codex, OpenCode, Copilot, or Pi and you keep losing context when you leave your desk, evaluate Paseo. The strongest reason is not the phone app. It is the combination of self-hosted agents, Git worktrees, provider configuration, supervised services, diffs, terminal output, and PR context in one remote-control layer.

If you only need a single local CLI agent, skip it for now. Paseo makes the most sense when agent work becomes parallel, branch-heavy, and annoying to supervise from one laptop.

Sources