Quickstart: your agent's first live plan

1. Connect the agent

• Open Settings and click to make a setup URL. It looks like https://taskhold.com/setup/<token>.md and expires after 3 days.
• Paste it into your agent with the words "set this up". The agent fetches it: the markdown holds a fresh API key, the config for that client, and the workflow instructions.
• Settings shows "Connected" when the agent's first call arrives.

To connect Claude Code by hand instead, run this with a key from Settings:

claude mcp add --transport http --scope user taskhold https://taskhold.com/api/mcp \
  --header "Authorization: Bearer API_KEY"

2. Open the live board

Open https://taskhold.com/dashboard/live in a browser tab you can see while the agent works. Click "Notify me" if you want a desktop notification when an agent gets blocked or finishes.

3. Give the agent a real task

Add a task in the web app (or let the agent file one), then say to the agent:

Take my next Taskhold task. Plan it in steps with taskhold_plan, report progress as you go, and block it with a question if you need me.

The card appears under Working now when the agent calls taskhold_start. Each step it plans becomes a segment of the bar; the segment fills as the step finishes, and the running step's message shows under the bar.

4. Make it the default

Add this block to CLAUDE.md or AGENTS.md so every session works this way without being asked:

## Tasks (Taskhold)

The person and I share a task list through the taskhold MCP tools. They watch a live board while I work.
- Session start: taskhold_get_current, then taskhold_updates for replies. Nothing in progress? taskhold_list status:todo and take the highest priority.
- taskhold_start before working on a task (agent: "Claude Code in <repo>").
- More than a couple of steps: taskhold_plan, then taskhold_start and taskhold_complete each step.
- taskhold_progress at milestones (N of M, or a percent) with a short message. Every few minutes at most.
- Need a decision: taskhold_block with one specific question, then check taskhold_updates.
- Done: taskhold_complete with what changed and how I checked it.
- A TODO or bug I'm not fixing now: taskhold_create (or taskhold_create_from_todo with file and line).

If you installed the Claude Code plugin from github.com/gettaskhold/taskhold-plugin, its skill carries the same workflow and you can skip the block.

Check that it worked

• The task shows a bar split into the steps the agent planned, with the agent's name on the card.
• The Activity feed lists "Started", the plan, progress messages and "Completed".
• The finished task moves to Finished recently with the agent's summary.

Nothing on the board? See /docs/troubleshooting.

What your agent sends, and how to disconnect it

Your agent sends task titles, steps, progress and messages to Taskhold. Taskhold never sends code or commands to your machine; your agent reads your replies only when it asks for them.

To cut an agent off, open Settings > API keys and click Revoke on its key. The agent's next call fails with 401; remove the Taskhold server from the agent's MCP config to stop it trying. Each machine has its own key, so revoking one leaves the others working.