tasuki quickstart
tasuki is a free task manager for people who work with AI agents. This page takes you from first sign-in to a working agent connection.
1. Sign in
Open https://app.usetasuki.com.
Sign in with Google, GitHub, or an email link.
After you sign in you will see four areas: Home/Today, Activity, Connect, and Settings.
2. Add your first task
Go to Home/Today.
Create a task the way you normally would. Keep the title short and concrete.
Quick-add text can carry hints such as #project, @label, p1, tomorrow, or every mon. tasuki shows those as chips you can untick before the task is saved.
Open Activity once so you know where changes land. Every change you or an agent makes shows up there, and you can undo it.
3. Connect your AI
Open Connect.
Create a personal access token. Copy it straight away. It is shown once.
When you create the token, pick a trust level (or Advanced scopes):
- Suggest only. The agent proposes changes. They land in a Review queue. You accept or reject each one.
- Trusted. The agent can add and edit tasks.
- Full. The agent can also complete and delete tasks. Everything is logged. Everything is undoable.
Start on Suggest only if you want a safety net. Move up when the workflow feels right.
Remote MCP URL:
https://app.usetasuki.com/mcp
REST API base:
https://app.usetasuki.com/api/v1
OpenAPI:
https://app.usetasuki.com/openapi.json
Both paths use a Bearer token unless your client uses OAuth.
Claude Code
claude mcp add --transport http tasuki https://app.usetasuki.com/mcp --header "Authorization: Bearer YOUR_TOKEN"
Claude.ai or ChatGPT
Add a custom connector. Use the MCP URL above. Those clients use OAuth.
Cursor and other MCP clients
{
"mcpServers": {
"tasuki": {
"url": "https://app.usetasuki.com/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOKEN"
}
}
}
}
Plain REST
List tasks:
curl -sS https://app.usetasuki.com/api/v1/tasks \
-H "Authorization: Bearer YOUR_TOKEN"
Add a task:
curl -sS -X POST https://app.usetasuki.com/api/v1/tasks \
-H "Authorization: Bearer YOUR_TOKEN" \
-H "Content-Type: application/json" \
-d '{"title":"Ship the quickstart"}'
4. What agents can do
Your agent can call MCP tools such as list_tasks, get_task, add_task, update_task, complete_task, delete_task, and list_activity.
What it is allowed to do still depends on the token trust level and scopes.
On add, update, complete, and delete, the agent can pass dry_run: true (or ?dry_run=true on REST). That checks scope, trust, and validation and writes nothing.
Review queue
On Suggest only, proposals wait in Review.
- Accept applies the suggestion as is.
- Edit then accept lets you change fields, then accept.
- Reject needs a reason the agent can read later (optional reason code plus free text).
5. Undo and Activity
Activity is the audit trail. Open it when something looks wrong and undo the change.
6. Export and delete
In Settings you can:
- download all your data as JSON
- delete your account
Privacy note: tasks are not used to train AI.
7. Troubleshooting
401 responses
The token is missing, wrong, or revoked. Create a new token on Connect and update the client header or connector.
Agent suggestion never shows up
You are probably on Suggest only. Check the Review queue and accept or reject the proposal.
Agent cannot complete or delete
That needs Full trust. Trusted can add and edit only.
Scope denied (403)
The token lacks the scope for that action. Create a new token with the right trust level or Advanced scopes.
Rate limited (429)
Wait for the window to reset, then retry. Response headers say when.
Nothing appears in the app after an agent call
Confirm the MCP URL or API base is exactly as above, and that the same account owns the token.
Need a clean slate
Use Activity to undo recent changes, or export from Settings before you delete anything important.
That is enough to run tasuki with an agent day to day. Keep Home/Today for your own list, Connect for tokens and trust, and Activity when you want the paper trail.