Skip to content

developdaly/botlanes

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

35 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

botlanes

Turn Claude Code or Gemini into a repeatable engineering pipeline.

board

The problem

You're shipping with Claude Code or Gemini. But your process looks like this: a bunch of terminals, a dozen open chats, mental notes about what's mid-review and what needs a follow-up. When the agent finishes, you forget to check. When you ask it a question, you're digging through scroll.

There's no system. It's just chaos.

The fix

botlanes is a Kanban board where every column is a real development stage — and when you drag a card into a column, an agent runs.

Not a chatbot. Not a one-shot prompt. A named gstack agent skill executing against your actual codebase, in the right directory, with the right context.

  Backlog → Office Hours → CEO Review → Eng Review → Design
                                                         ↓
              Ship ← QA ← Code Review ← Implementation ←┘

You see every task. Every task has a stage. Every stage runs the right agent. When the agent needs you, the card lights up. When it's done, you drag it forward.

This is what agentic development looks like when it has a process.


Prerequisites

  • Bun — runtime and package manager
  • One of (or both) AI CLIs:
    • Claude Code CLI — the claude binary must be in your $PATH
    • Gemini CLI — the gemini binary must be in your $PATH
  • gstack — provides the agent skills that power each column

Install

git clone https://github.com/developdaly/botlanes
cd botlanes

Start

bun run src/cli.ts start

The server starts on a random port. The URL is printed on startup — open it in your browser.


How it works

1. Create a project

Point botlanes at a directory — your repo. Every card in that project runs its agent with that directory as the working directory.

2. Write a card

A card is a task. Give it a title and description. Be specific — the agent reads this as its brief.

3. Drag to run

Drag a card from Backlog into any stage column. botlanes spawns a claude or gemini subprocess (depending on which one you selected for the project) running the matching gstack skill. The card turns blue: running.

4. Watch, respond, move

Open the card to watch live output. If the agent asks you something, the card turns amber and waits. Answer in the UI. When the stage finishes, drag the card forward.

5. Attach Multimedia

Need to provide more context? Drag and drop images, videos, documents, or audio directly onto a card. You can also paste screenshots with ⌘V. These attachments are automatically passed to the agent as multi-modal context during execution.

modal-review modal-complete

The pipeline

Stage What the agent does
Backlog Parked — nothing runs
Office Hours YC-style product brainstorm — clarify what you're actually building and why
Autoplan Run the full planning sequence automatically
CEO Review Scope and strategy check — is this the right thing to build?
Eng Review Architecture and implementation plan review
Design Visual design and UI decisions
Implementation The agent writes the code
Code Review Pre-landing diff review
QA Automated QA testing and bug fixes
Security Audit CSO-style security review and vulnerability scan
Ship PR creation and merge
Visual Review Designer's eye QA on the live app

Run stages in any order. Skip what you don't need. Run the same stage twice. You're in control — botlanes just makes sure the right agent runs when you say so.


Why this matters

Most AI-assisted development is stateless. You fire off a prompt, get output, context disappears. There's no thread connecting your idea to your shipped code.

botlanes gives you that thread. Each card carries its full history — every stage output, every question answered, every decision made. You can see at a glance where every task is and what the agent did to get it there.

It doesn't replace your judgment. It just removes the overhead of managing it.


Architecture

Layer Tech
Backend Bun + TypeScript (src/server.ts)
Frontend Preact + htm — no build step, ES modules via esm.sh
Styling Tailwind CSS via CDN
State File-based, stored in .gstack/
Agent runner Bun.spawnclaude or gemini CLI subprocess

State is local and file-based. No database, no cloud, no accounts. Your tasks stay on your machine.


Commands

bun run src/cli.ts start   # start the server
bun test                   # run tests
bunx tsc --noEmit          # typecheck

License

MIT

About

Drag a card into a column, an agent runs.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors