Pad CLI
A project tracker on SQLite in a single binary: CLI, web UI and a /pad skill through which an agent manages tasks, plans and ideas
- 207 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The agent creates and edits tasks, plans and documents in your workspace through the CLI and MCP
- The local Pad server listens on `localhost:7777` while it runs
- Moving to Pad Cloud sends project data to an external service
Install
- Works in
Manual install
brew install PerpetualSoftware/tap/padThe first quick-start command: then `pad init` in the project folder, `pad server open` and `/pad onboard` in an agent session.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent creates and edits tasks, plans and documents in your workspace through the CLI and MCP
- The local Pad server listens on `localhost:7777` while it runs
- Moving to Pad Cloud sends project data to an external service
This is third-party code. Review the repository files before installing.
What it does
Pad stores a project's tasks, ideas, plans and documents in a local SQLite database and opens them in three ways: CLI commands, a web UI at `localhost:7777` and a `/pad` skill for the agent. The `pad init` command works out what is needed (setup, sign-in, workspace, skill installation) and is safe to run again. The agent gets typed collections with fields, conventions and playbooks that explain how the project runs. `/pad onboard` starts an interview, and with shell access the agent also inspects the codebase. For Claude Desktop, Cursor, Windsurf, Claude Code and Codex there is an MCP server: `pad mcp install <client>` registers it and the client reads workspace state and calls tools. The skill works in Claude Code, Cursor, Windsurf, Codex, OpenCode, GitHub Copilot, Amazon Q and JetBrains Junie.
Who it is for. Solo developers and small teams who want to keep tasks next to code on their own machine and talk to an agent about the project, without tickets in someone else's cloud.
Good fit when
- An agent needs a shared source of truth for tasks, plans and ideas that outlives sessions
- You want to mark done work and create tasks with `/pad` without leaving the chat
- Work goes on offline and the tracker must work fully offline
- Later the project will move to the cloud for a team and an export is needed
Not a fit when
- A large team with roles and approval is needed: the local version is for one person and a team needs the separate cloud one
- Tasks already live in Linear or Jira and there is no plan to move them
- Installing a binary on the work machine is not possible
Example request
/pad create a task to add rate limitingLimitations
The README describes installation through Homebrew (`brew install PerpetualSoftware/tap/pad`), and for other OSes see the project releases. Local mode works without sign-up. Sync, invites and roles exist only in Pad Cloud at `app.getpad.dev`, and moving goes through export and import. The `/pad` skill and MCP server are installed separately with `pad init` and `pad mcp install`. The repository is large, about 4.7 thousand files. The latest commit in the repository is from November 2026.
How to disable. Stop the Pad server and remove the binary with `brew uninstall pad`. The `/pad` skill and MCP entries are removed from the skills directory and client configuration, and the SQLite database sits in the Pad data directory and is deleted by hand.
FAQ
How do I install Pad?
Run `brew install PerpetualSoftware/tap/pad`, then `pad init` in the project folder. The web UI opens with `pad server open` at `localhost:7777`.
Which agents does it work with?
The `/pad` skill supports Claude Code, Cursor, Windsurf, Codex, OpenCode, GitHub Copilot, Amazon Q and JetBrains Junie. The MCP server installs for Claude Desktop, Cursor, Windsurf, Claude Code and Codex.
Where is the data?
In a local SQLite database on your computer. Pad Cloud is needed only if you move a project yourself and invite a team.
What does /pad onboard do?
It starts the built-in onboard playbook: the agent runs an interview, inspects the codebase with shell access and adapts collections, conventions, roles and playbooks to the project.
How do I connect MCP?
With `pad mcp install claude-desktop` or `cursor`, `windsurf`, `claude-code`, `codex`, `--all`. After a client restart the `pad` server appears in the list.
Discussion
No questions or reviews yet.
Related
CLI toolsPlanDB is an issue tracker for AI agents: a dependency graph, atomic task claiming, notes and critical path analysis in SQLite.
A scaffold for Claude Code and Codex: tasks in files, developer and QA roles, manual approval and shared state.
A terminal kanban board in Python that Claude Code and other agents can read and change through a CLI or MCP
Eight Claude Code skills for Obsidian: Granola notes, NotebookLM imports, TaskNotes tasks, recall of past sessions, chat export
A skill that keeps project memory in plain Markdown files: session logs, a wiki, rules, a to-do list and a note on where to resume
Local memory for Claude Code, Cursor and Codex in one SQLite file that feeds past decisions and lessons into the agent's context
README in short
The README shows a three-command quick start, a comparison with Linear, Jira and Notion, features for developers and agents, MCP installation and a description of collections, conventions and playbooks. The cloud version is covered separately.