che-cli

A small collection of CLI utilities, glued together by one dispatcher binary (che). Each tool is a subcommand. No plugin system, no manifest — drop a script in lib/che/<topic>/, add a case arm in bin/che, done.

Source: github.com/chevp/che-cli

Pages: installation · workflows · architecture · troubleshooting

Install

git clone https://github.com/chevp/che-cli.git
cd che-cli
./install.sh
che doctor

Default install prefix is ~/.local. Override with PREFIX=/usr/local ./install.sh.

Requirements: bash, git, curl, python3 (or python). On Windows, run from Git Bash or WSL.

Commands

`che commit`

Stages all changes, generates a commit message via the configured LLM provider, asks for confirmation, commits, optionally pushes.

che commit
che commit --push
che commit --dry-run
che commit --edit

Provider defaults to Ollama (local). Switch via CHE_PROVIDER:

Provider	Selector	Auth
Ollama	`CHE_PROVIDER=ollama`	none (local server)
Claude Code	`CHE_PROVIDER=claude-code`	handled by the `claude` CLI (subscription / login)
GitHub Copilot	`CHE_PROVIDER=copilot`	handled by the `copilot` CLI (Copilot subscription)

che never handles API keys directly. Cloud LLMs are reached only through their official CLIs — install claude or copilot and let it own auth. To remove the local fallback and force every prompt through Claude Code, set CHE_FORCE_CLAUDE_CODE=1.

The HTTP request shape Ollama speaks is kept as an ad-hoc REST sample under client/http/ — useful for poking at the local server in VS Code REST Client or JetBrains HTTP Client without going through the CLI.

For local Ollama setup, see the companion guide: cura-llm-local.

`che issue`

Drafts a GitHub issue via the configured LLM, prints it for review, and opens it through gh issue create. Subcommands cover listing and closing.

che issue "race condition in workflow loader"   # AI drafts title + body, asks to confirm
che issue create --dry-run                      # draft from current branch + diff, do not open
che issue list --limit 5                        # list open issues for this repo
che issue close 42 --reason "fixed in v1.4"     # close issue with comment

Requires gh installed and authenticated (gh auth login).

The drafted issue body always begins with a small YAML frontmatter block:

---
status: open            # open | in-progress | blocked
progress: 0%            # free-form, e.g. "0%" or "0/3 steps"
---

<rest of the body>

che issue list and che status parse that block and render the same status badge as plans — so you can track progress finer than GitHub’s open/closed without inventing labels. Edit the frontmatter directly in the issue body to update status. GitHub’s closed state always wins: closed issues never appear in the list, regardless of frontmatter.

`che status`

One-screen overview of the current repo + che-cli configuration:

active provider, model, reachability
git: branch, upstream, ahead/behind, working-tree state, short status
submodules with sync state
last 5 commits
issues — top 5 open GitHub issues for this repo (via gh), with a status badge parsed from the issue body’s YAML frontmatter (when present)
pull requests — top 5 open PRs with draft / review-decision state
plans — local .che/plans/*.md files with status badge (open / in-progress / done / blocked)

Issues, plans, and (after refactor) anything else that carries a frontmatter block share a single parser at lib/che/frontmatter.sh and render through one common frontmatter_status_badge helper — add a new state once, it shows up everywhere.

che status            # full overview, including GitHub + plans
che status --short    # one-line summary, no network calls

GitHub sections are silently skipped if gh is missing, unauthenticated, or the repo has no GitHub remote. Plan files use a small YAML frontmatter (name, status, optional progress); see .che/plans/README.md for the format.

`che run` / `che workflow`

Runs a YAML pipeline of shell-script steps from .che/workflows/<name>.yml. The workflow file declares order, args, and inputs — never inline bash — so your existing scripts/ stay executable on their own.

che workflow list                       # discover workflows in this repo
che workflow show <name>                # print parsed step plan
che run <name> --tag=v1.2.3             # execute (alias: che workflow run)
che run <name> --tag=v1.2.3 --dry-run   # plan only

Full guide: workflows.

`che doctor`

Verifies dependencies and providers, with platform-aware install hints.

che doctor               # all checks
che doctor provider      # only the active provider
che doctor ollama        # ollama install + reachability + model
che doctor docker        # docker install + daemon running
che doctor workflow      # python + PyYAML for che workflow / che run

Architecture

bin/che                         dispatcher (~30 lines)
lib/che/
  platform.sh                   OS detection: darwin | windows | wsl | linux
  provider.sh                   routes calls to the active provider
  git/        check.sh, commit.sh
  ollama/      check.sh, client.sh        <p>_ping / _generate / _has_model
  claude-code/ check.sh, client.sh        wraps the `claude` CLI binary
  copilot/     check.sh, client.sh        wraps the `copilot` CLI binary
  docker/      check.sh, client.sh
  workflow/   check.sh, loader.sh, list.sh, show.sh, run.sh
  workflow.sh                   `che workflow` sub-dispatcher
  doctor.sh                     `che doctor` — runs all checks

Each <provider>/client.sh exposes the same three functions, which provider.sh routes to based on $CHE_PROVIDER. To support a new provider: drop a folder, implement those three functions, register it in provider.sh.

cura-llm-local — local LLM server setup (Ollama)
chevp-mcp-suite — MCP server collection
chevp-blender-mcp — Blender MCP integration
quon-cli — companion CLI tooling

Roadmap

che-cli is meant to grow organically. Candidates:

che pr — open a PR with an LLM-generated title and body
che review — local LLM code review on staged changes
che branch — name a new branch from a short intent description

che-cli

Install

Commands

che commit

che issue

che status

che run / che workflow

che doctor