Skip to main content
Back

Manus Turns Its AI Agent Into a Persistent Chief of Staff

Manus ships native MCP connectors, reusable Skills, and recurring runs that stay in the same context, turning the agent into an always-on personal Chief of Staff.

Repo, Full Access
  • Manus launched Scheduled Tasks 2.0 with native MCP connectors and a Chief of Staff template.
  • Recurring runs can now continue in the same task, preserving context and history across executions.
  • OAuth-based connectors include Gmail, Slack, Notion, Calendar, Drive, Granola, Asana, and Fireflies.
  • Two importable Skills ship at launch: Daily Assistant for briefings and Personal CRM for relationship tracking.
  • Trusted workflows can skip confirmations and send, post, or publish without manual approval each time.
  • Scheduled Tasks 2.0 is available to all users; Manus is now part of Meta.

Manus has rolled out a package of updates that reframe its agent as a persistent operator rather than a one-shot chatbot. The headline pieces are Scheduled Tasks 2.0, a new set of native MCP connectors, and a Chief of Staff template that stitches them together. The pitch is straightforward: hook up your work tools once, teach Manus your preferences, and let it run recurring jobs against the same evolving context.

The Chief of Staff workflow is built on three primitives. According to Manus, you connect tools, encode your preferences as Skills, and put prompts on a schedule. MCP connectors are prebuilt integrations that let Manus read data from an app, perform actions within it, and coordinate workflows across multiple platforms from a single natural language prompt. Authentication runs through OAuth 2.0, the same method used by major platforms, with the user controlling exactly what Manus can access.

What is actually new under the hood

Scheduled Tasks 2.0 is the more interesting half of this release, because it changes how recurring agent work behaves. Run options let users choose whether each run continues in the same task or starts as a separate task. That sounds small, but it is the difference between an agent that wakes up amnesic every morning and one that remembers yesterday's open threads.

A few other controls round it out:

  • Skip confirmations lets trusted workflows proceed without asking for approval before sending, publishing, or posting.
  • Connectors let a scheduled task use connected apps as relevant data sources.
  • For work organized in Projects, scheduled tasks can reuse the shared setup already defined there, including files, skills, connectors, instructions, and output standards.
  • Web apps built with Manus can now include scheduled actions of their own, useful when an app needs to refresh data, run a script, update a dashboard, send a reminder, or generate a recurring summary.

The connector roster covers most of the obvious productivity surface area: Gmail, Google Calendar, Google Drive, Slack, Notion, Outlook, Box, Dropbox, Asana, Todoist, monday.com, Calendly, Cal.com, Superhuman, Granola, and Fireflies. No API keys, no config files.

Two ready-made skills to start with

Manus is shipping two importable Skill bundles so you do not have to design the workflow from scratch:

  • Daily Assistant handles morning briefs, evening recaps, email triage, and persistent task tracking. It checks calendar events, triages email, updates a to-do list, and sends summaries through Slack and email.
  • Personal CRM scans Gmail and Calendar to maintain contact profiles, log interactions, calculate relationship scores, and produce weekly follow-up recommendations with draft outreach emails.

Both can be pulled in via the import-skills flow on Manus and edited like any other prompt artifact.

Meeting recap dashboard showing completed tasks checklist, action items table with owners and due dates, and a prompt to draft follow-ups

Why context-preserving schedules matter

The real shift here is treating a schedule as a continuation of a task rather than a fresh invocation. The schedule follows the place where the work lives, not just the time on the calendar. If you have wired up an agent loop on top of an LLM API before, you know that recreating context every run is where most of the brittleness lives, half the engineering work is reloading state and reconciling what changed since the last run.

It is also where competitors are stumbling. Claude Code users have been filing bugs reporting that cloud scheduled tasks do not have access to MCP connectors, the agent runs ToolSearch for each connector, gets no matching deferred tools, and stops to ask the user to connect MCPs that are already connected. Manus is explicitly designing around that failure mode by treating connectors as first-class scheduled-run citizens.

Practical patterns to try

The most useful example prompts from the launch and docs are concrete enough to copy directly:

  • Every weekday at 7:30 AM, generate a brief with the schedule, overdue tasks, and priority action items; at 6:00 PM, send what got done, what is open, and tomorrow's preview.
  • After every meeting, draft follow-ups in your voice using your saved tone Skill, then weekly compile a contact spreadsheet and flag relationships that have gone cold.
  • Every Monday, update the customer feedback summary in this Project using the files and format already here.

Availability and the Meta angle

Scheduled Tasks is now available to all users, and in any task, Project, or web app built with Manus where recurring work belongs, you can tell Manus what you want scheduled and it will keep running in the right place. Worth noting in the background: Manus is now part of Meta, which makes the push toward always-on personal agents read less like a startup experiment and more like positioning for a broader consumer agent play.

The honest limitations are still the ones every agent product faces. Skills only encode preferences you actually write down, so a sloppy tone description produces sloppy follow-ups. Skip confirmations is powerful and dangerous in the same breath, and any workflow that touches outbound email or Slack deserves a few weeks of supervised runs before you take the training wheels off.