Adeptly CLI
A local web interface over Claude Code: drafts plans with feature recipes, runs them as an agent crew, checks them against git.
- 67 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The Crew runs headless `claude` sessions in your repository, edits files and can open a pull request.
- The `--tunnel` mode lets anyone with the address control your `claude`.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/adeptlyDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The Crew runs headless `claude` sessions in your repository, edits files and can open a pull request.
- The `--tunnel` mode lets anyone with the address control your `claude`.
This is third-party code. Review the repository files before installing.
What it does
Adeptly starts with `npx adeptly` and opens a page on `localhost:3000`. Plans live in `docs/plans/` of the current directory, and every AI step calls the local `claude`: there is no backend, no keys and no telemetry. For each plan section the app adds a recipe: which Claude Code subagents, skills, hooks, MCP servers and modes to use. The Crew runs the plan through a chain of roles from architect and builder to reviewer, security check and pull request. The Verify tab compares the plan file list with the real git diff, and Usage shows token spend parsed from transcripts in `~/.claude`.
Who it is for. Teams of several developers on one repository where everyone drives Claude Code and wants to see other plans before the pull request lands.
Good fit when
- A large task needs a plan that says where subagents and hooks fit.
- Two Claude Code sessions edit one repository and must be split by files.
- After a session the question is whether the agent left the plan or deleted tests.
- Claude Code spend grew and a breakdown by model and day is wanted.
Not a fit when
- Claude Code is not installed: the app works only on top of a local `claude`.
- A shared server with accounts for the whole team is required.
- A single short prompt would solve the task without a plan.
Example request
Draft a plan for adding CSV export, name the Claude Code subagents and hooks it needs, and run the Crew in approval mode.Limitations
A working, signed-in `claude` is required: every AI step spends your limits. Cost estimates use public rates and can differ from the bill, and Adeptly infers window caps from the times you hit a limit. `adeptly --tunnel` exposes the app through a Cloudflare quick tunnel: anyone with the address drives your `claude`, so enable it knowingly. The Crew runs agents in your repository and changes files.
How to disable. Stop the `npx adeptly` process and delete `docs/plans/` and `.adeptly/` if the plans and usage log are no longer needed. The app installs no global package or service.
FAQ
Where does the code go?
Only where it goes during normal `claude` use. The web interface and server run on your machine, and Adeptly has no account, analytics or backend.
Where are plans stored?
In `docs/plans/` as markdown. The team syncs them through Git, and there is no separate store.
What does the Verify tab show?
Done and not-done plan steps, files outside the plan and red flags such as deleted tests, removed assertions and added `.skip` or `.only`.
Can it be exposed externally?
Yes, through `adeptly --tunnel`. The README warns that anyone with the URL can drive your `claude`, so Git is the better route for team work.
Which Claude Code versions are required?
The README names no minimum version: an installed, signed-in `claude` is needed, and Adeptly calls it for every AI step. Run `claude --version` before the first launch.
Discussion
No questions or reviews yet.
Related
CLI toolsA personal ~/.claude for Claude Code: twelve subagents, review commands, worktree and goal skills and a Go status line.
Claudette runs several Claude Code sessions in parallel, each in its own git worktree, from a window or the console
A personal Claude Code setup: four loading layers, an advisor model mode, ten hooks and working commands
Multi-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
A CLI that runs parallel Claude Code sessions from your plan, each in its own git worktree, with stage checks and a knowledge base
A Claude Code template: 15 department agents, five skills and an approval queue for outward actions such as emails and invoices.
README in short
The app is a thin interface over the local `claude`. Its main bet is recipes that point out which Claude Code features to use in each plan section.