iloom CLI
iloom runs each Claude Code task in its own worktree and saves plans and decisions as comments in your issue tracker
- 111 ★ on the repository
- Last commit
Extended access
Why this level
- Creates worktrees and branches and commits, merges or opens a pull request on `il finish`
- Writes comments to the issue tracker and keeps a Linear or Jira API token in its settings
- In `--yolo` mode it runs Claude without confirmation prompts
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/iloomDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Creates worktrees and branches and commits, merges or opens a pull request on `il finish`
- Writes comments to the issue tracker and keeps a Linear or Jira API token in its settings
- In `--yolo` mode it runs Claude without confirmation prompts
This is third-party code. Review the repository files before installing.
What it does
iloom installs as the npm package `@iloom/cli` and works on top of Claude Code. `il start 25` takes an issue from GitHub, Linear or Jira and creates a loom for it: a separate git worktree, its own database branch (Neon is supported), its own port and its own environment files. Web apps get a port derived from a base number; for CLI projects iloom builds an isolated copy of the binary, so versions from two tasks run side by side. Inside a loom a chain of agents does the work. The Enhancer turns a one-line request into requirements with acceptance criteria, the Evaluator rates complexity and picks a simple or complex route, the Analyzer reads the code and libraries before any edit, the Planner writes a plan that marks which steps can run in parallel, and the Implementer writes the code. iloom stores the analysis, plan and decisions as comments on the issue, so a teammate can see why the agent chose a given structure. For a large task, `il plan` splits it into child issues with dependencies, and swarm mode launches parallel agents, one worktree each. `il finish` runs tests and lint, commits, resolves conflicts and merges or opens a pull request. The VS Code extension installs with `il vscode` and shows decisions, assumptions and open questions in a recap panel.
Who it is for. Teams where Claude Code writes large pull requests, the reviewer needs to see the agent's reasoning and a developer runs several tasks in parallel.
Good fit when
- An agent sent a 2,000-line pull request and the reviewer needs the assumptions and reasons in the issue itself
- You work on three tasks at once and do not want to stash and switch branches
- An epic is split into subtasks with dependencies and parallel agents suit them
- Every branch should get its own Neon test database so migrations never collide
- The team lives in Linear or Jira and wants the agent's plan there, not in a chat
Not a fit when
- You are on native Windows, which is unsupported and needs WSL
- You have no Claude CLI or subscription, which iloom relies on
- The task takes five minutes and a separate environment with a plan only slows the edit
- There is no issue tracker and you want to keep agent context in the repository itself
Example request
Take issue 25 from our GitHub, post a plan in the comments and start implementation in a separate loomLimitations
- The product is at an early stage, as the author says in the README. Native Windows is unsupported; a separate WSL guide exists.
- Needs Node.js 16 or newer, Git 2.5 or newer and an installed Claude CLI. The author recommends a Claude Max subscription because iloom runs through it.
- The license is BSL-1.1 (Business Source License), which is outside the list of OSI open source licenses. Commercial use terms are in `LICENSE`.
- Database isolation works with Neon only, and Linear or Jira need an API token in the iloom settings.
How to disable. `il cleanup` removes a loom together with its database branch without merging. `npm uninstall -g @iloom/cli` removes the tool itself. `il telemetry off` turns off anonymous telemetry.
FAQ
How do I install iloom for Claude Code?
Run `npm install -g @iloom/cli`, then `gh auth login`, and start `il init` in the project folder. The first task starts with `il start 25`, where 25 is the issue number in the tracker.
What is a loom in iloom?
A loom is the isolated environment of one task: a separate git worktree, database branch, port and `.env` files. Inside a loom shell (`il shell`) the variables `ILOOM_LOOM` and `ILOOM_COLOR_HEX` are set.
Where does iloom save the agent's plans and decisions?
In comments on the issue in GitHub, Linear or Jira. The VS Code extension installed by `il vscode` shows the same material in a recap panel.
What does the --yolo flag do?
It is shorthand for `--one-shot=bypassPermissions` and runs the whole pipeline without confirmation prompts. By default iloom asks for approval at each stage: enhance, plan, implement.
Which issue trackers does iloom support?
GitHub by default, with login through `gh auth login`. Linear connects through the `il init` wizard with an API token, and Jira (Atlassian Cloud) is configured in `.iloom/settings.json`.
Discussion
No questions or reviews yet.
Related
CLI toolsXylocopa is a visual orchestrator for Claude Code agents: GTD tasks, parallel agents in git worktrees, phone approvals.
A Claude Code working configuration: a planner with per-milestone review, sub-agents and skills for analysis, refactoring and docs sync
A plugin for Claude Code: you read one spec and parallel agents ship it to a commit with green tests.
Terminal coding agent with profiles, a swarm of subagents, checkpoints and code improvement cycles scored on six metrics
A workflow template for Claude Code and five more agents: a backlog task, a diff review and a commit.
A coding agent for small, cheap LLMs: splits a goal into a task tree with contracts and verifies each leaf with builds and tests.
README in short
The README opens with the question of trusting code an AI wrote and explains the idea of keeping the agent's reasoning in the tracker. It continues with a quick start, a description of the agents, a table of `il` commands, setup through the `il init` wizard and sections on swarm mode, Docker and telemetry. It ends with requirements and limitations.