Task management for your AI agents

Write plain markdown. Get a kanban board. Let AI agents manage it alongside you.

Open source·No vendor lock-in·Your data, your git repo
kanban
waiting for markdown...
How it works

Write markdown. See structure.

Every heading becomes a column. Every subheading becomes a card. Metadata like @assignee, #label, and priority:high become visual badges.

sprint-23.md
# Sprint 23
## Implement auth flow
@alice #backend priority:high
- [x] Design JWT schema
- [ ] Build login endpoint
- [ ] Add refresh token logic
## Update dashboard
@bob #frontend est:6h
- [ ] New chart component
- [ ] Filter sidebar
kanban view
Sprint 23 2
Implement auth flow
#backend 1/3
A
Update dashboard
#frontend 6h
B
Why markdown

The format AI was trained on.

Most tools force AI to work through clunky APIs and rigid schemas. AutoMD stores everything as plain markdown — the one format every language model already understands deeply.

AI reads it natively

Every LLM is trained on markdown. It's the format AI understands best — no adapters, no translation layers, no lossy conversions. Your agents read and write tasks in their native tongue.

It's just files

No proprietary database. No vendor lock-in. Your tasks are .md files in a folder. Move them, back them up, grep them, version control them. They're yours forever.

Human and machine readable

Markdown is the rare format that's equally comfortable for humans and AI. You read it in your editor. Agents parse it in their context window. Same file, no impedance mismatch.

No per-seat licensing

Running 5 AI agents in Jira or Linear? That's 5 seats you're paying for. Here, agents just read and write markdown files — no seats, no API limits, no vendor tax. Sync everything through git.

Views

Three views. One source of truth.

Switch between editor, checklist, and kanban. They all read and write the same .md file. No sync issues. No database migrations. Just text.

app.automd.ai/project
# Sprint 23
## Implement auth flow
@alice #backend priority:high
- [x] Design JWT schema
- [ ] Build login endpoint
- [ ] Add refresh token logic
## Update dashboard
@bob #frontend est:6h
- [ ] New chart component
Sprint 23
Implement auth flow #backend
A
Design JWT schema
Build login endpoint
Add refresh token logic
Update dashboard #frontend 6h
B
Sprint 23 2
Implement auth flow
#backend
A
Update dashboard
#frontend 6h
B
MCP Integration

Your AI teammates show up to standup.

Connect Claude, GPT, or any MCP-compatible agent. They read your board, understand context, and take action — moving cards, setting priorities, adding estimates. 20+ MCP tools for full board control.

Claude via MCP connected
Y
You

"Triage the backlog and assign tasks for this sprint based on team capacity"

C
Claude

Looking at the backlog...

Moved "Auth flow" to Sprint 23 → @alice
Set priority:high (blocking deployment)
Moved "Dashboard update"@bob
Added est:6h based on similar past tasks
Left "Refactor utils" in Backlog (low priority)
3 tasks assigned · 14h estimated · Sprint starts Mon
Triage & Assign
Auto-sort your backlog by priority and assign to available team members
Daily Standup
Summarize progress across all boards and flag blockers
Sprint Planning
Allocate tasks based on team capacity and velocity
Features

Built for how you actually work.

Real-time Sync
WebSocket updates. Every edit, every move, instantly everywhere.
Keyboard-First
Command palette, vim motions, shortcuts for everything. Your mouse can rest.
Dark Mode
System-aware. Warm amber on dark. Your eyes will thank you at 2am.
Inline Metadata
@assignee #label priority:high due:friday — all parsed from plain text.
Drag & Drop
Natural kanban reordering. Grab a card, drop it where it belongs.
Git-Friendly
Commit, branch, merge, diff your tasks. Version control for project management.

Get Started

Choose how you want to run AutoMD. Either way, your data stays yours.

Self-Host

Run it on your own server. Single Docker container, zero external dependencies.

View Setup ↓
Coming Soon

Cloud

We'll run it for you. Same markdown files, same MCP tools, zero DevOps.

Coming Soon

Your data stays portable. Export to markdown anytime. No lock-in, ever.

Self-Host

Your server. Your data. Two commands.

AutoMD runs in a single container. No database to manage, no Redis to configure. Your tasks live as markdown files in a directory you control.

terminal
$ docker compose up -d
automd-server Started
Ready at http://localhost:4800
# That's it. No database. No Redis.
# Just your markdown files.
Single Docker container, zero external dependencies
First user becomes admin automatically
Optional S3/R2 backup for peace of mind
Update: docker pull && docker compose up -d

Open source. Built in public.

AutoMD is source-available under the Sustainable Use License. Read the code, run it yourself, contribute back.