Set up

Use your own Vikunja

Already run Vikunja? Docket works with any Vikunja 2.x server. Your existing projects become Docket lists, and nothing about your server has to change.

#Android app

Sign in with your server's address (for example https://tasks.example.com) and your username and password. Docket makes a token for the phone; the password isn't stored.

#Windows app and CLI

Install Docket with the Windows installer, then connect it to your server instead of a local one:

PowerShell
& "$env:LOCALAPPDATA\Programs\Docket\setup-tokens.ps1" -ServerUrl https://your-vikunja

It signs in once and makes two tokens: one for the app and CLI, and a no-delete one for the AI connector. You can also enter the address and an API token by hand in Settings > Connection, or with docket login --server https://your-vikunja.

If you don't want a local server at all, run the installer with -NoScheduledTasks -SkipAccountSetup, then run setup-tokens.ps1 as above.

#AI connector

The connector needs your server's address and a token. docket-mcp setup makes a no-delete token for you:

Shell
export DOCKET_VIKUNJA_URL=https://your-vikunja
export DOCKET_DATA_DIR=~/.config/docket
docket-mcp setup        # asks for your username and password once, saves the token
docket-mcp              # local (stdio), for Claude Desktop, Claude Code, LM Studio…
docket-mcp --http       # remote (HTTP), for claude.ai and ChatGPT behind your HTTPS proxy

For remote use, set DOCKET_MCP_KEY (24+ characters) and DOCKET_PUBLIC_URL (your public HTTPS address) too, or run the container image instead:

Shell
docker run -d --name docket-connector -p 127.0.0.1:8080:8080 \
  -e DOCKET_VIKUNJA_URL=https://your-vikunja -e DOCKET_MCP_KEY=... -e DOCKET_PUBLIC_URL=https://your-vikunja \
  -v docket-connector:/data ghcr.io/niteshaggarwal/docket-connector
docker exec -it docket-connector docket-mcp setup

Then route these paths from your reverse proxy to the connector, and everything else to Vikunja (or Docket web): /mcp, /oauth/*, /.well-known/oauth-protected-resource*, /.well-known/oauth-authorization-server*, /api/tools/* and /openapi.json. deploy/docker/Caddyfile is a working example.

All connector settings are listed in how AI works with Docket.

#Docket web

To give your Vikunja the Docket look, run docket-web in front of it: see Docket web.

#Good to know

  • Docket stores its features in plain Vikunja: My Day is a label, Important is a favourite, steps are a checklist in the description. See how Docket maps onto Vikunja.
  • Avoid renaming the "My Day" label or the labels that start with "↻"; Docket uses them.
Edit this page on GitHub