User repository on GitHub
Run AI coding agents in parallel, each in its own git worktree. Desktop app for Linux, macOS and Windows (Tauri + Rust).
- Forks
- 1
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
CommunityJev Agents & MCP
Yardsort
GPL-3.0 desktop app for Linux, macOS and Windows that gives every coding-agent task its own git worktree and its own terminal, with live diffs and optional Jev review of changed files.
- Stars
- 10
- Language
- Rust
- License
- GPL-3.0
- Published by
- Community
- Author
- joaoh82
- Last verified
- Oct 2026
About this project
Yardsort Run AI coding agents in parallel — each on its own track. Yardsort is an open-source desktop app for Linux, macOS and Windows. Every task gets its own git worktree and its own terminal, running the coding agent of your choice. yardsort.sh · Download · Quick start · Documentation · Contributing · Contact. A sorting yard is where rail cars are sorted onto parallel tracks and later joined back into one train. That is the job: fan work out onto parallel branches, then merge it back.
Quoted from the project's README. Read the full README.
- Languages
- Rust
Highlights
- Workspaces are ordinary git worktrees and branches, so you can inspect or undo anything with git; worktrees made elsewhere can be imported.
- Agents run in a real PTY with their own interface and Yardsort never parses their output, and terminals live in a background process.
- Assist asks Jev small typed questions about each changed diff and badges files that look unrelated to the task, or that add a secret or weaken a test.
- Deleting or archiving a workspace always keeps the branch, and uncommitted changes are never discarded without a second explicit confirmation.
Watch out
GPL-3.0, and our README copy is cut off inside the scriptable-tools bullet, so the ys CLI surface, install steps, system requirements and the dedicated Assist page are not documented in what we read. Cross-platform support is stated as a requirement rather than a measurement: Linux, macOS and Windows are described as first-class and built and tested together on every change. Assist is off unless you bring your own TypeSafe API key, and the README publishes no accuracy or false-positive figures for the badges it produces. Network use is limited to a version check you can switch off, plus Jev calls only when Assist is on with your key; agents still use their own logins, and the GitHub integration goes through the gh CLI, so Yardsort holds no forge credential of its own. The activity record only notes when each agent started in a workspace and how it ended, and it never reads a word the agent printed; attributing a changed file is an inference from the agent reporting it, from a command that was running when it was last written, or from nobody the record knows. The Pull requests and Tasks panels are GitHub-shaped and everything goes through gh; without gh your browser opens the forge's own form, on GitHub, GitLab, Bitbucket or Gitea alike. Cost figures are labelled estimates, and the Codex plan limits shown are that agent's last reported values. Project memory is gated on your approval: agents propose lessons and only what you approve reaches an agent. Outcomes are your own words, with a merged pull request counting as kept until you say otherwise. The author positions this against Conductor and Superset, naming cross-platform support as the thing they do not meet. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
The fan-out is the automatic part; the merge path is a panel rather than a keyboard shortcut. You commit what an agent wrote, push it and open the pull request from the panel you reviewed it in, with the title written from the commits. With the GitHub CLI the pull request opens from there and its number and check results land on the workspace row, and a conflicting one can be handed back to the agent that opened it to resolve. The consolidated Pull requests view can start a workspace on one and merge, close or reopen it, each confirmed first and naming whose it is.
No. Agents run in a real PTY with their own interface, and the README is emphatic that Yardsort never parses their output, so whatever an agent can do in your terminal it can do here. The activity record is separate and opt-in: it notes when each agent started in a workspace and how it ended, even if that happened while the window was closed, and it gets that detail from agents reporting their own tool calls, file changes and turns as metadata, with their settings untouched.
It asks Jev small typed questions about each changed diff and badges files that look unrelated to the task, or that add a secret, weaken a test or switch a check off. It can also suggest a harness and an effort for the message you are typing. It is off unless you supply your own TypeSafe API key, and the README publishes no accuracy or false-positive rates, so treat a badge as a prompt to look, not a verdict.
Terminals live in a small background process, so agents keep working while the window is closed or after it crashes, and reopening repaints every screen where it got to. Closing with work in flight asks first. Deleting or archiving a workspace always keeps the branch, and uncommitted changes are never discarded without a second, explicit confirmation. Because workspaces are ordinary worktrees and branches, you can also inspect or undo anything with git yourself.
GPL-3.0, and it is a desktop app for Linux, macOS and Windows built and tested together on every change.
Terminals live in a small background process, so agents keep working while Yardsort is closed or after it crashes; reopening repaints every screen where it got to, and closing with work in flight asks first.
It compares Claude Code, Codex and Grok by day, model and workspace, shows estimated API costs and Codex's last reported plan limits, and inspects live CPU and memory for every terminal's process tree.
More Jev resources like this
35GitHub stars
Rust terminal tool that asks Jev noul, choice and score questions about piped text or files and returns calibrated probabilities with stable exit codes for shell scripts, CI jobs and agents.
Jev Agents & MCP#community#jevbest#catalog-import
Community
shaharia-lab/jev-cliRust
shaharia-lab/jev-cliRust5GitHub stars
Grok Build wrapper that asks Jev one noul per tool before the first model call of a turn, then holds the shortened tool list so the prompt cache survives.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
daniel-farina/nitroRust
daniel-farina/nitroRust2GitHub stars
Claude Code plugin for Rust review that runs Clippy and cargo-semver-checks itself, has Jev triage changed units for defects no tool reports, then verifies each candidate finding before it reaches you.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
kindintelligence/jev-rust-reviewRust
kindintelligence/jev-rust-reviewRust51GitHub stars
Local LinkedIn recruiting workspace on Browser Use's Jev Ultrafast: Jev screens titles, drives navigation, and assesses profile excerpts against your brief while a live browser stays visible; matches and evidence export as JSON.
Jev Agents & MCP#community#jevbest#catalog-import
Community
skeptrunedev/jev-recruiterPython
skeptrunedev/jev-recruiterPython33GitHub stars
Codex Skill plus stdlib-only Python CLI that recommends which installed skill to read next for a task, allowing no-skill or uncertain answers and never executing candidates.
Jev Agents & MCP#community#jevbest#catalog-import
Community
win4r/jev-skill-suggesterPython
win4r/jev-skill-suggesterPython72GitHub stars
MCP server over stdio that gives coding agents fourteen Jev judgment tools, turning evidence plus an enumerable question into auto, review, or escalate actions from calibrated probabilities.
Jev Agents & MCP#community#jevbest#catalog-import
Community
PyModel/jev-judge-mcpPython
PyModel/jev-judge-mcpPythonFrom the Jev community
Posts from builders shipping with Jev right now.