npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

branchling

v0.4.0

Published

Git-native task execution for people and coding agents with verification-gated closure.

Readme

test npm license: MIT

A Git-native backlog for work done by people and coding agents.

Tasks are markdown files in your repository. branchling hands them out one at a time, records who claimed each one and why it changed, and refuses to close a task until its own verification commands pass:

plan → claim → execute → capture discoveries → verify → review

  • A queue. branchling next chooses a task and reserves it in one act, so two sessions asking at the same moment get two different tasks.
  • A contract. branchling done runs the task's verification: commands and refuses to close it if they fail. There is no --force.
  • A ledger. Every change records who made it, through what, and why.
  • An executable plan. plan.yaml orders work in dependency-aware waves.
  • Plain files. Views, indexes and the browser page are computed from the files and never committed. The task state travels with its branch and goes through code review beside the code it describes.
branchling run --agent "claude -p @{task_file}" --max-attempts 2

That is the whole loop — next → your agent → done — until the queue is empty. run starts exactly the command you supply; a task that fails its contract twice is parked with the reason, never closed.

What it is not

It is not a tracker for people outside the repository. There is no hosted board, login or notification; only people with a clone can see the backlog. A generated viewer.html snapshot is read-only, and a viewer served on loopback is the only viewer mode permitted to write.

It is not a managed agent factory. It does not contain, choose, host or schedule agents for you. branchling next --json and branchling done are a queue and a gate that any orchestrator, CI job or shell loop can drive.

| | External tracker | GitHub Issues | branchling | |---|---|---|---| | State follows the branch, and git revert takes it back | ✗ | ✗ | ✓ | | A task is readable with no login and no network | ✗ | ✗ | ✓ | | An AI agent edits a task with the same tools it edits code | ✗ | partly | ✓ | | The diff of a task shows up in code review | ✗ | ✗ | ✓ |

Getting it

npm install -g branchling
branchling --version

To try it without installing anything, npx branchling --help runs the same package once and leaves nothing behind.

Requires Node 18 or newer; no runtime dependencies. CI runs Linux and macOS on Node 18, 20 and 22, and Windows on Node 22.

Quick start

branchling init --dir ./backlog                     # config, template, one example task
branchling new --title "Write a README for a new user" --priority P1
branchling query --status pending --priority P0,P1  # ask; reads the task files directly
branchling next --actor agent:claude                # take the next task, reserved
branchling show TASK-2                              # deployment, other branches, blockers
branchling done TASK-2                              # runs its verification, or refuses
branchling                                          # open the viewer in your browser

init never overwrites an existing file. After it, every command finds the backlog by itself from anywhere inside the repository. It also points your CLAUDE.md or AGENTS.md at branchling instructions overview; --skills adds the agent skill, and branchling doctor names whatever is still missing. An unknown command or flag fails rather than doing nothing, and branchling <command> --help prints the flags of each one.

A task file

---
id: TASK-2
title: "Write a README for a new user"
board: main
priority: P1
status: pending
owner: unassigned
blocked_by: []              # ids that must be closed before this one starts
verification:               # how to check the task is actually done
  - id: links
    bash: "branchling check --docs"
---

## Goal
## Context
## Acceptance criteria
- [ ] Every link in the README resolves. [proof: links]

The tool ticks a criterion after its verification entry passes, so a box ticked by hand is a claim rather than evidence. Statuses, priorities and boards are your project's vocabulary, declared in config.yaml.

See it work

Each scenario runs offline in a throwaway repository:

mkdir /tmp/try && cd /tmp/try && git init .
branchling init --dir backlog --no-example
branchling new --dir backlog --title "Work that outlives its session"
branchling new --dir backlog --title "A second piece of work"

One agent stops, another continues from the repository. A passer-by cannot clear somebody else's claim; the owner's successor is briefed from the files alone.

branchling take TASK-1 --dir backlog --actor agent:first
branchling release TASK-1 --dir backlog --actor agent:second --reason "not mine"
# ✗ TASK-1 is in_progress, owner: agent:first
branchling resume TASK-1 --dir backlog --actor agent:first
branchling release TASK-1 --dir backlog --actor agent:first --reason "the session that held it ended"

Two agents ask at once, and get different work. With a plan.yaml, --plan also keeps work scheduled for a later wave from starting early.

branchling next --dir backlog --actor agent:a           # TASK-1
branchling next --dir backlog --actor agent:b           # TASK-2, never TASK-1
branchling next --dir backlog --plan --actor agent:c

A finished task that is not finished is refused. A task with no verification cannot be closed at all, and one whose verification fails is left byte-for-byte untouched.

branchling done TASK-1 --dir backlog
# ✗ TASK-1 has no closing contract

The full script, with recorded output, is docs/demo/scenario.md.

Working with AI agents

Every task is self-contained: the context, the files to read first and the way to verify the result are in the file, not in a conversation. The procedure for claiming, working and closing a task is printed by the tool, in your project's vocabulary:

branchling instructions overview
branchling skills install      # a skill pointer for editors that load skills
branchling mcp                 # the same commands over the Model Context Protocol

skills install never overwrites an existing file; skills install --refresh replaces one only when it is a version the package shipped, so an edit of yours is never lost. Local profiles connect any provider through an adapter you own; see the guide.

Documentation

  • The guide — the task file, the queue and agent profiles, where the data lives, composition with --json, importing from GitHub Issues, pull-request summaries and audits.
  • The manual — statuses, configuration, plan.yaml, seeding, the dispatcher, guards, the viewer and the --json contract.
  • docs/branchling-global-tool.md — the four rules every other design decision follows from.
  • Building on top of it — the stable contract for a separate interface: the --json kinds, the writing commands, version pinning, and where the built-in viewer stops.
  • LINEAGE.md — where this came from. The tool tracks itself with itself: its own backlog is in backlog/.

Contributing

Contributions are welcome under the MIT licence with a DCO sign-off (git commit -s); there is no CLA. Read CONTRIBUTING.md before opening a pull request, and SECURITY.md before reporting a vulnerability.

License

MIT. The branchling name and logo are not covered by it; see docs/brand/.