Now in Agent Mode

Two modes. One task workspace.

Taskitos works in User Mode and Agent Mode. You capture and manage tasks the way you always have — and now your AI agents like Claude Code and Codex can too, writing to the same durable task memory. Every agent action waits for your review.

Connect your agent See how it works

User Mode

Everything you already love — quick add, voice capture, smart categories, stars, tags, and persistent reminders. You stay in the driver's seat.

Agent Mode

Your AI coding agents connect securely over MCP and drop follow-up work straight into Taskitos — tagged, attributed, and queued for your review.

Don't let good ideas die in a transcript

You're deep in a session with Claude Code or Codex and you realize there's a follow-up worth doing later. Normally it gets buried in the chat or lost in a generic to-do list. With Agent Mode, the agent saves it to Taskitos — and Taskitos remembers, reminds, and nags until it's done.

The agent captures the work

You say "add this to Taskitos for next week." The agent calls create_task with a title, due date, and tags like taskitos and Codex.

Taskitos remembers it durably

The task is stored with its due date, source context, and agent attribution — outside the transcript, in your real task memory.

Taskitos follows up

The task flows through the same reminders, smart nagging, and auto-rescheduling as everything else, so it never quietly disappears.

You review and decide

It shows up in your Agent tab for a quick accept, edit, or dismiss. Nothing an agent does is final until you say so.

You stay in control: the Agent tab

Agent-created tasks aren't a separate universe — they appear in your normal date views based on when they're due. But every create or update from an agent is also marked pending review and surfaced in a dedicated Agent tab, with clear provenance like “Created by Codex via MCP.”

Keep it as-is

Edit

Adjust title, date, or tags

Dismiss

Drop it if it's not needed

What agents can — and can't — do

create_task

Capture a new task with an optional due date, tags, and the context it came from.

list_tasks

Read your open tasks back at the start of a session so the agent picks up where you left off.

update_task

Mark a task complete or extend its due date. Limited on purpose — no arbitrary edits.

undo_created_task

Roll back a task the same agent just created, before you have acted on it.

Narrow by design

An agent can read your tasks, create new ones, mark tasks complete, and extend due dates. It cannot delete arbitrary tasks, change billing, access admin data, or bypass Taskitos permissions. Access is granted through OAuth consent and you can revoke it anytime.

How to connect your agent

Add the Taskitos MCP server to your agent

Point your agent at https://taskitos.com/mcp. In Claude Code that's claude mcp add taskitos https://taskitos.com/mcp; in Codex, add it as an MCP server in your config.

Sign in and grant consent

Your agent opens a browser window to connect Taskitos. Sign in, review what the agent is allowed to do, and approve. No personal access tokens to copy around.

Start capturing work

Ask your agent to add, list, or complete Taskitos tasks in plain language. Each action shows up in your Agent tab for review.

Revoke anytime

Connected agents and their permissions live in your account settings. Disconnect any agent whenever you want.

Try saying things like

Add "ship the agent-mode page" to Taskitos for next week.

Pull my open taskitos tasks so we can pick up where we left off.

Mark the landing-page task as done.

Built on the Model Context Protocol

Agent Mode is a hosted MCP server with OAuth, so any MCP-compatible agent can connect — not just Claude Code and Codex. It discovers authorization automatically through standard OAuth protected-resource metadata.

Let your agents remember the follow-ups

Create your Taskitos account, connect your AI agent, and stop losing good ideas to the scrollback.

Get started free Explore all features