Skip to content

Personal wire-up checklist

Use this page as the short operational checklist after you have installed the Hub. For detailed options, follow the linked guides instead of trying to configure everything here.

Day in the life

  1. Hub running — use the installation guide for Docker Compose (recommended), docker run, or source/uv setup.
  2. Connect once — prefer the Connect one-liner. Copy it from Settings → Coding agents and do not export the server token into your shell.

If your MCP client has Auth / Authenticate, configure ADHD_HUB_PUBLIC_URL, leave OAuth enabled, then approve the browser prompt. Static Bearer auth and CLI connect remain supported. See MCP Auth / Authenticate.

# macOS / Linux / WSL
curl -fsSL http://127.0.0.1:8787/install.sh | sh -s -- /path/to/project

# Windows PowerShell
irm http://127.0.0.1:8787/install.ps1 | iex

adhd-hub doctor --project /path/to/project

After a Hub upgrade:

  • run adhd-hub doctor --project /path/to/project on each active project;
  • use adhd-hub setup . --refresh for stale managed guidance;
  • add --install-skills only when you intentionally want global skill updates.

See project agent setup.

  1. Skills (global) if you skipped --skills on connect:
# Install from GitHub (recommended):
npx skills add uniskela/adhd-hub -g
# Or, while developing an unreleased local checkout:
cd C:\path\to\adhd-hub
npx skills add ./skills -g
  1. Optional third-party companions (i-have-adhd, Superpowers, Ponytail, Graphify, Context7, agent-browser, RTK, Serena, Humanizer) — toggle under Settings → Coding agents, or see Recommended third-party companions.
  2. Start a coding session in any project — skip Hub tools for trivial/read-only questions and tiny edits. For substantial work: once per meaningful session resolve_project → session_digest. If resuming a known thread, reuse its thread_id; otherwise use check_overlap before potentially new/duplicate work and reuse only when the Goal matches. See project agent setup.
  3. Pause or finish — checkpoint with upsert_progress(thread_id=...) using goal, focus (one action), ≤3 next_steps, blocked_reason only if blocked, and resume_step. Prefer completion.ready when present: unfinished work → pause_thread; complete Goal → mark_done (do not retry mark_done if rejected for remaining work). See project agent setup and ADHD-friendly writing and planning.
  4. Review — open http://127.0.0.1:8787/ui, filter by project, check Stale, open Gitea issues if board sync is on.
  5. Indexer backstop (optional daily): see the advanced indexer schedule. It is intentionally not part of the main docs navigation because most users do not need it.

Optional OpenClaw check-in

If OpenClaw is part of your setup, install the same skills there:

npx skills add uniskela/adhd-hub -g -a openclaw

Then pair from Settings → Phone alerts (recommended): Start pairing, paste the prompt into OpenClaw, and Approve. Manual webhook/token (and .env) remains available. Full steps: OpenClaw connection and alerts.

Local smoke (Windows)

cd C:\path\to\adhd-hub
copy .env.example .env
# Edit .env: ADHD_HUB_AUTH_TOKEN, optional ADHD_HUB_PUBLIC_URL
# (and ADHD_HUB_OAUTH_ENABLED=false only if you need Bearer-only rollback)
copy .cursor\mcp.json.example .cursor\mcp.json   # if present
uv run adhd-hub serve --host 127.0.0.1 --port 8787

Or Docker:

docker compose up -d --build

Cursor hooks (optional)

See adapters/cursor-hooks.md — sessionStart/stop prompts that nudge digest + progress writes.

Deploy to lab

.\scripts\sync-and-deploy.ps1 -HostName root@<tailscale-host-or-ip> -RemoteDir /opt/adhd-hub

Then set MCP URL to http://<tailscale-host-or-ip>:8787/mcp and ADHD_HUB_PUBLIC_URL to that base for README → /ui links. See Homelab deployment for the full Proxmox/Tailscale flow.