Show HN: Smith – AI Agent Orchestrator
A developer has launched Smith, a free AI agent orchestrator that lets programmers run Claude Code, Codex, Gemini CLI, Aider, and OpenCode side by side in isolated git worktrees. It works like a terminal multiplexer for AI coding tools, and it points to a fast-growing need for be...
According to the project listing at getsmith.dev, submitted to Hacker News by the developer under the handle netnameus, Smith is a multi-agent command center currently at version 0.1.0. The tool is free for up to 3 projects, with a beta tier priced at $5 and major releases at $24. It runs natively on macOS and Windows, is built with Tauri and Rust, and ships as a binary weighing roughly 25 megabytes.
Why This Matters
The AI coding assistant market now has at least 5 serious terminal-based agents, and developers who want to run more than one are stuck cobbling together manual workflows. Smith is one of the first tools to treat that coordination problem as a first-class product. The free tier makes adoption nearly frictionless, and the $24 ceiling for major releases keeps it accessible to solo developers and small teams alike. If it delivers on its promise of git worktree isolation per agent, it could save teams hours every week in merge conflict resolution alone.
Daily briefing from 50+ sources. Free, 5-minute read.
The Full Story
Managing a single AI coding agent is already a context-switching challenge. Managing three or four simultaneously, across separate branches, on different machines, has until now required considerable custom scripting or accepting messy git histories. Smith positions itself as the answer to that specific problem, built from the ground up for developers who want to run parallel AI coding sessions without the chaos.
The core concept is close to what tmux or iTerm2's split panes do for terminal sessions, but extended for the specific demands of AI coding tools. Smith opens Claude Code, Codex, Gemini CLI, Aider, and OpenCode in separate panes, each with custom naming and live status subtitles. Output streams in real time. An MCP (Model Context Protocol) server built into Smith lets agents query the orchestrator's state, and a smith.toml configuration file means teams can share an identical setup across machines without manually recreating . The git worktree isolation feature is arguably the most practically significant detail. Each agent gets its own worktree on its own branch, which means two agents can edit the same codebase simultaneously without stepping on each other. A dedicated diff review panel, opened with Command+Shift+D, shows a file tree with per-file insertion and deletion counts. Developers can stage or discard changes per file and commit directly to the base branch without leaving Smith. The mock terminal output shown on the site demonstrates an agent refactoring an authentication module from session cookies to JWT tokens, producing 42 insertions and 18 deletions across 3 files, which is exactly the kind of task where worktree isolation prevents a second agent from interfering mid-edit.
Remote agent support is another standout feature. Developers can point an agent at a remote server by adding a host entry to the smith.toml config, and Smith wraps the session in tmux for resilience. It inherits the developer's existing ~/.ssh/config, agent, and known_hosts files, and supports password authentication with a test-connection button in the settings modal. Full PTY streaming means the experience feels native even over a remote connection, and MCP tools function over remote sessions too.
Mobile access rounds out the feature set in a way most developer tools still skip. Smith runs a built-in web server on port 3200, meaning any device on the same network can open a browser and get full terminal streaming, sidebar navigation, and agent controls. On small screens the sidebar auto-hides, making triage from a phone genuinely functional rather than theoretical.
Key Details
- Smith is currently at version 0.1.0 and is free for up to 3 projects.
- Paid tiers are priced at $5 for the beta and $24 for major releases.
- The binary is approximately 25MB and is built with Tauri and Rust.
- Supported agents at launch include Claude Code, Codex, Gemini CLI, Aider, and OpenCode.
- The Hacker News submission received 2 points and 1 comment at the time of posting.
- Git worktrees are stored in a .smith-worktrees directory with one subdirectory per agent.
- The built-in web server runs on port 3200 with WebSocket terminal streaming.
- Remote agent sessions are wrapped in tmux for connection resilience.
- macOS and Windows are supported at launch, with Linux listed as coming soon.
- The Command+K shortcut filters the view to only agents requiring attention, sorted by last activity.
What's Next
Smith's immediate roadmap priority appears to be Linux support, which the site lists as coming soon, and that addition would bring the tool to the servers where most remote development work actually runs. Watch for the MCP server integration to deepen over time, since that protocol is becoming a de facto standard for agent-to-tool communication and broader support there would make Smith a natural hub for any MCP-compatible agent that emerges. If the project gains traction on Hacker News beyond the initial quiet launch, expect a community-driven push for more agent integrations beyond the current 5.
How This Compares
The closest conceptual predecessor to Smith is not another AI tool but tmux itself, and that comparison matters. tmux solved the problem of persistent, multiplexed terminal sessions in 2007 and became infrastructure developers depend on daily. Smith is attempting the same thing for AI agents, and the design decisions reflect that ambition. The keyboard-driven interface, the split-pane layout, and the configuration-as-code approach through smith.toml all borrow from the tmux playbook deliberately.
Compare Smith to Anthropic's own Claude Code, which handles a single agent session well but offers no native mechanism for running parallel agents on isolated branches. Codex and Aider have similar single-agent designs. There are AI tools that offer agent orchestration at the API level, including LangChain and CrewAI, but those require developers to write coordination logic in Python. Smith's value proposition is that it handles orchestration through a graphical interface without requiring custom code. That is a different audience and a different abstraction layer.
A related Hacker News project posted roughly 59 days before Smith's submission demonstrated AI agents designing and shipping a complete application called NinjaFlix in 36 hours at a total cost of $270. That project showed what's possible when agents can work in parallel on real codebases. Smith is, in effect, the tool that makes that kind of parallel workflow accessible without building custom infrastructure from scratch. The $270 NinjaFlix build was notable precisely because it was novel. Smith is trying to make that kind of workflow routine.
The minimal initial engagement on Hacker News, 2 points and 1 comment, should not be read as a verdict on the tool's quality. Developer infrastructure tools frequently launch quietly and grow through word of mouth once a few power users integrate them into daily workflows. Check the latest AI news for follow-up coverage as adoption data emerges.
FAQ
Q: What is an AI agent orchestrator and why do developers need one? A: An AI agent orchestrator is a tool that manages multiple AI systems working at the same time. Developers need one because running two AI coding assistants on the same codebase simultaneously, without an orchestrator, causes conflicting edits and messy git histories. Smith handles that coordination automatically by giving each agent its own isolated branch and worktree.
Q: Is Smith free to use for personal projects? A: Smith is free for up to 3 projects with no time limit. Developers who need more than 3 projects can pay $5 during the beta period, and major releases are priced at $24. The tool runs on macOS and Windows, with Linux support listed as coming in a future release.
Q: How does Smith compare to just opening multiple terminal tabs manually? A: Opening multiple terminals manually gives you no coordination between agents, no worktree isolation, and no unified diff review. Smith adds git branch isolation per agent, a built-in diff panel for staging and merging changes, native OS notifications when an agent needs input, and a mobile-accessible web interface on port 3200, none of which you get from raw terminal tabs. Read the Smith setup guide for a walkthrough of the full workflow.
Smith is a small tool solving a real problem that is only going to grow as AI coding assistants multiply and developers start running more than one at a time. Whether it becomes the standard orchestration layer for AI-assisted development depends on whether the team can keep pace with the quickly evolving agent ecosystem. Subscribe to the AI Agents Daily weekly newsletter for daily updates on AI agents, tools, and automation.
Get stories like this daily
Free briefing. Curated from 50+ sources. 5-minute read every morning.
