Agent-native issue tracking
for ambitious Claude Code users

curl -fsSL https://setup.fp.dev/install.sh | sh -s
Get Started Explore Workflows

v0.10.0

Why this?

agents love CLIs

fp uses a CLI to manage issues for your project. Claude helped design it, and Claude loves using it.

context across sessions

Compaction doesn't cut it. fp lets you clear sessions safely. Claude progressively loads relevant context for a task in each new session.

local-first

Issues, comments, and context are stored locally on your machine. No external services. No MCP server.

smooth local code review

See what code changed and why in a lightweight local interface. fp collects your comments and sends them straight back to Claude.

Installation

1

Install the CLI

curl -fsSL https://setup.fp.dev/install.sh | sh -s
2

Add the Claude Code plugin

fp setup claude
3

Initialize in your project

fp init

Workflows

Tell Claude to use fp and it handles the rest:

claude code — my-project
> use fp to break down adding auth
◆ Created 4 issues:
auth-wgzx Add user authentication ├─ auth-fohq Set up auth provider ├─ auth-cbae Create login page └─ auth-zowy Add session middleware
> looks good, pick up auth-f
◆ Starting auth-fohq. Reading issue context...

Then run ! fp review in Claude to open the code review UI and send comments back to Claude:

claude code — my-project
> ! fp review
Review UI for working copy at http://localhost:3435/review/wc
◆ Opening review UI. Leave comments on the diff and I'll address them.
fp review UI showing a diff view with inline commenting
The local review UI. Comments get batched for Claude to fix.

Submit the review, and your comments appear in the terminal. Claude can take it from there:

claude code — my-project
> [Review submitted with 1 comment]
◆ You have absolutely reviewed! Fixing the Drizzle client.

Once you're done, move to the next task with a clean slate to prevent context rot. Run /clear, and tell Claude to start the next task:

claude code — my-project
> /clear
(no content)
> use fp. tackle the next task in the auth epic
◆ Starting auth-cbae. Reading issue context...

Commands

Claude can drive all task management with the CLI, but for us humans, these commands are also helpful:

fp tree [issue-id] Display issue hierarchy as ASCII tree fp issue list --status todo List issues in table format, filtered by status fp issue show <issue-id> Show detailed info including comments fp review <issue-id> Open web-based diff review UI (omit issue-id to review working copy)

Run fp help to see all the other commands.

Motivation

fp makes you and Claude work better. It plays to the strengths of coding agents, and helps avoid their most harmful quirks. It is built on a few simple observations:

Agents (and humans) like focused tasks. Breaking work into bite-sized tasks helps both Claude and us write better code. Keeping those tasks outside of the coding session makes context handoff between sessions smooth and easy.

Agents love CLIs. CLIs are easy for Claude to explore and learn. We built a simple CLI, and Claude was our first user. Whenever Claude hallucinated a command or a flag, we added it. By design, Claude is an fp power-user.

Agents are litterbugs. Claude scatters one-off markdown files all over your codebase, and it's annoying. What Claude really wants is a scratchpad for its research and work. fp gives it a place to take notes without littering CODEBASE_EXPLORATION_V2_ALT.md in your repo.

Humans need a solid review tool. We wanted an interface to capture feedback on Claude's code, and wrap it all up into a single message. GitHub is too slow. Hence, fp review. You shouldn't have to manually copy-paste notes after reviewing diffs in your IDE or terminal.

fp is your shared brain with Claude. It's your record of what you did, what you're doing, and what you want to do. It's simple, effective, and once you use it, it's hard to imagine working without it.