Paseo

Paseo is a self-hosted interface for running coding agents across desktop, mobile, web, and command-line clients. The project is designed for developers who already use agent CLIs such as Claude Code, Codex, GitHub Copilot CLI, OpenCode, or Pi and want one control surface for starting, monitoring, and coordinating those agents. Its core architecture is a local daemon that manages agent processes on the user’s own machine. The self-hosted design is the most important part of Paseo. Agents run in the user’s local development environment, with local repositories, credentials, tools, configuration, and skills. That means a developer can keep execution close to the machine where the code and permissions already live. Clients such as the desktop app, mobile app, web app, and CLI connect to the daemon, so a task can be started from a desk and checked later from a phone. Paseo is not trying to be a new foundation model. It is orchestration software for existing coding agents. The README describes parallel agent runs, live output attachment, follow-up messages, and remote daemon workflows. The CLI examples show commands for running tasks with Claude or Codex providers, listing running agents, attaching to a session, and sending additional instructions. Paseo also ships skills for handoffs, loops, advisors, and committee-style analysis. The repository is open source under AGPL-3.0, but users still need the underlying agent CLIs and credentials. In practice, that means model or subscription costs come from Claude Code, Codex, Copilot, OpenCode, Pi, or whatever provider is connected. Paseo’s role is to coordinate those tools and make them reachable from multiple devices, not to replace their billing or setup requirements. Paseo is best for developers and engineering teams experimenting with multi-agent coding workflows. It can help when agents need to run in parallel, when work should continue on a local workstation, or when a developer wants to supervise tasks remotely. It is less useful for nontechnical users or teams that want a fully hosted coding platform. The value is local control, cross-device access, and practical orchestration for agentic software development. For OpenTools readers, the first evaluation should focus on security and fit with existing development habits. Install the daemon on a noncritical machine, connect one provider, run a narrow task, and inspect how logs, credentials, worktrees, and follow-up messages are handled. Teams using remote access should review network exposure before leaving a daemon reachable from other devices. The listing should be reviewed again when the project publishes new pricing, setup, or security documentation. Until then, the safest approach is to treat the official website or repository as the source of truth, validate claims before production use, and start with a narrow pilot that measures whether the tool saves real workflow time.