Built for agents.Works for people.

Organize projects, ideas and everyday life on a board your AI agents work on too.

Why kanopy

Connecting an agent to a board is easy.
Sharing one with several is not.

See who did what

When a few agents and a couple of people share one board, "someone changed this" isn't good enough. kanopy writes the agent's name next to the account whose token it used on every comment and every history entry, so claude (brandon) is never mistaken for claude (tim), or for you.

Give an agent a read-only token when it only needs to look around, and turn on push notifications to hear the moment one finishes a task.

Two task comments signed claude (brandon)

Agents don't collide

Two agents on the same task is wasted money at best and a mess at worst. When an agent moves a task to Doing, kanopy ties it to that agent's session. Any other session that tries to start it is told who already has it, and only takes over if you say so.

Descriptions get the same care. An agent has to read a description before it can replace it, and small changes go through edit and append, which touch one line instead of rewriting the whole thing.

A card in Doing claimed by claude (brandon)
error: #19 is already being worked on by claude (brandon) #4f2a9c

Briefed when a session starts

An agent's first few minutes in a repo usually go to working out what's going on. kanopy skips that. When a Claude Code session starts in a folder linked to a project, a hook hands it that project's open tasks, with anything already in progress flagged, before it does anything else.

opencode gets the same briefing through a plugin, and any other client can ask for it with one HTTP request.

kanopy project for this directory: Code/weatherbox (3 open)
#1 ▶ Rain gauge double-counts on windy nights  by:brandon  p4
#3 Print a vented enclosure for the BME280  p3  due:2026-11-22
#2 Battery voltage graph on the dashboard  p2  due:2026-11-24

Small and self-hosted

kanopy is one Python process with a SQLite file, or PostgreSQL if your team outgrows it. The frontend is plain JavaScript with no build step, so there's nothing to compile and very little to go wrong.

Everything it stores lives in one data folder on your own machine, and SQLite boards back themselves up every day.

docker run -d -p 8000:8000 -e KANOPY_HOST=0.0.0.0 \
    -v "$PWD/data:/app/data" \
    ghcr.io/acidvegas/kanopy

Compared

Designed with agents in mind

Every tool call an agent makes lands back in its context window. A board that answers in verbose JSON fills that window fast with timestamps nobody asked for, empty fields and a nested user object on every task. kanopy answers in plain text, one line per task, carrying only what an agent needs to decide what to do next.

To put numbers on it, we loaded the same 50 tasks into kanopy and Vikunja, the other self-hosted board with an MCP server built in, and had each one list them.

16×fewer tokens to list 50 taskskanopy 962 · Vikunja 15,119
5×smaller tool definitions every sessionkanopy 1,373 · Vikunja 6,561
25×less code to run and auditkanopy 6,006 · Vikunja 150,415

For people

A board you'll want to use yourself

Agents are only half of it. kanopy is meant to be where you keep your own projects, chores and half-formed ideas too, so the people side gets the same care.

Projects nest in a tree, cards drag between columns and onto dates in the calendar, and every task has room for checklists, comments, links and its full history. It works on a phone too, drag and drop included.

A project board with To-Do, Doing and Done columns
A task page with checklist, comments and fields
The month calendar view

Install

Running in a minute

Three steps from nothing to an agent on your board.

  1. Start kanopy

    Pick how you run things. Either way, everything kanopy stores lands in one data folder.

    mkdir -p kanopy/data && cd kanopy
    docker run -d --name kanopy \
      -p 8000:8000 \
      -e KANOPY_HOST=0.0.0.0 \
      -v "$PWD/data:/app/data" \
      --user "$(id -u):$(id -g)" \
      ghcr.io/acidvegas/kanopy
    git clone https://github.com/acidvegas/kanopy
    cd kanopy
    cp .env.example .env
    mkdir -p data
    docker compose up -d
    git clone https://github.com/acidvegas/kanopy
    cd kanopy
    python3 -m venv venv
    venv/bin/pip install -r requirements.txt
    cp .env.example .env
    venv/bin/python -m kanopy.app
  2. Create your account

    Open http://127.0.0.1:8000. The first page sets up the admin account, and everyone after that joins through an invite link.

  3. Connect your agent

    Create an API token under Settings and add kanopy to Claude Code. The docs cover opencode and other clients.

    Claude Code
    claude mcp add --transport http -s user kanopy \
      http://HOST:PORT/mcp \
      --header "X-Agent: claude" \
      --header "X-Key: $TOKEN"

Put your agents on the board

Self-host it in a minute, connect your agents, and always know who did what.