You built the whole thing with Claude in a weekend.
Now it's Tuesday and you can't remember what you shipped.

The project board that works
with your AI agent

For solo builders and tiny teams who ship with AI. Connect via MCP — your agent reads the board, picks up tasks, logs its work, and never loses context between sessions.

Get Early Access See how it works

Backlog 4

KD-087

Email notification preferences

enhancement

KD-091

Bulk task import from CSV

feature

Spec 2

KD-084

Sprint velocity dashboard

ui P1

In Progress 3

KD-079

Command palette search

frontend

KD-082

MCP token logging

api

Testing 1

KD-076

Backlog decay badges

frontend

Deployed 5

KD-074

Capture flow redesign

core

Sound familiar?

Monday's context is gone by Wednesday

You vibe-coded three features. Now you're burning tokens re-explaining what you already built because the AI doesn't remember last session.

Your co-founder's AI doesn't know what yours did

Two people, two agents, zero shared state. You're Slacking each other screenshots of terminal output to stay in sync.

You have no idea what anything cost

You shipped fast, but how many tokens did that auth system take? Which feature was cheap and which blew your budget? No clue.

How it works

Add one MCP connection.
Your agent does the rest.

No setup guides. No config files. Connect and go.

Your agent already knows the project

Kanbang serves your project context, priorities, and working conventions through MCP. Open a new Claude session on Friday and it picks up exactly where Monday left off — no "here's what I've been working on" preamble.

→ mcp get_project("kanbang")

prefix: KD, tier: advanced

modules: [ui, api, mcp, core]

workflow: inbox → backlog → spec →

in-progress → testing → deployed

→ mcp get_project_backlog()

12 tasks, 3 stale, 1 cold

✓ Agent knows the project

You never write docs. They write themselves.

Every AI session captures what changed, why, and what it cost. Three weeks later when you wonder "why did we build it this way?" — the answer is already there. No Notion pages. No commit archaeology.

→ mcp update_task(KD-084, status: testing)

Added sprint velocity chart

Files: app.js +142, style.css +38

→ mcp log_tokens(KD-084)

input: 12,847 · output: 8,421

Task cost: $0.31

✓ Logged to effort timeline

Two builders, two agents, one brain

Invite your co-founder. Their AI agent sees the same board yours does. When their agent ships a feature and moves it to testing, your agent knows instantly. No Slack, no standups, no "hey what are you working on?"

→ mcp list_tasks(status: "in-progress")

KD-079 Command palette search [you]

KD-082 MCP token logging [alex]

· websocket: task:updated KD-082

· alex's agent moved KD-082 → testing

→ mcp get_task("KD-083")

Next highest priority: KD-083

✓ Picking up next task

Capabilities

Lightweight structure. Zero overhead.

Works with any MCP agent

Claude Code, Codex, Gemini, Cursor — anything that speaks MCP connects to Kanbang out of the box.

Context that survives sessions

Your agent reads the board on startup. No more pasting "here's what we did yesterday" into every new chat.

Know what each feature costs

Token spend per task, per sprint, per project. See where your AI budget actually goes.

Bring a co-founder, not a process

Shared board, shared state. Their agent sees your work, yours sees theirs. No standups required.

Capture from anywhere

Telegram bot, command palette, or just tell your agent. Ideas go to the inbox. Decide what's real later.

Docs that write themselves

Wiki, diagrams, decision logs. Generated from agent activity, not from you opening Notion at midnight.

Where this is going

You decide what to build. The agents build it.

Move a card to "In Progress" and your AI agent picks it up automatically — reads the spec, writes the code, logs what it did. When it's done, your co-founder's agent runs the tests.

You're the product owner. The board is the factory floor. The AI agents are the workers. We're not there yet — but every piece of Kanbang is built to make this real.

The project board that workswith your AI agent