Introducing MCP connections, scheduled tasks, and shared project files in Cenmax AI

Today, we’re introducing three additions to Cenmax AI: MCP connections, scheduled tasks, and shared project files. Together, they let you build workflows that produce useful outputs, such as a maintained keyword database, a weekly research briefing, a queue of article drafts, or a repository review saved to your local folder.

You define the outcome, choose the tools and files the task can use, and decide when it runs.

What’s launching

MCP connections on web and desktop

Connect remote MCP servers to give Cenmax AI access to capabilities offered by other services. Depending on the server, these can include keyword research, documentation search, content management, publishing, and other actions.

Choose Allow & connect, and Cenmax AI discovers and activates the server’s tools automatically. For services that support OAuth, sign in through your browser. For services that require an API key, enter it in the secure credentials field. Credentials stay outside the model’s chat context.

You select which servers each chat or scheduled task can use. Approving a connection covers all its tools, including tools the server adds later; it does not select that server for every conversation or task.

Scheduled tasks

Run a task once, hourly, daily, or weekly. Specify its instructions, select the model and MCP connections, and choose whether it can update shared project files.

Cloud tasks can run while your desktop app is closed. In Cenmax AI Desktop, you can also create local-only tasks that run within a selected folder while the app is open. You can authorize file writes when creating a local task so it can save its work unattended.

Shared project files

Keep working files inside a project so different chats and cloud tasks can read and update the same material.

A research task can add rows to a CSV. A writing task can read approved rows and prepare drafts. A later task can update their status. File revisions let you inspect how the work has changed.

The examples below show how these additions can work together. Actions performed through an external service depend on the capabilities and permissions of its MCP server.

1. Turn keyword research into a content pipeline

Outcome: A maintained topic database and a queue of article drafts based on approved opportunities.

Add a keyword CSV to a project and connect a compatible keyword research service. Schedule a research task to find relevant opportunities, record supporting data, and avoid duplicate entries.

Find new keyword opportunities for our audience. Add relevant topics to keywords.csv with their search intent, available metrics, source, research date, and a short explanation of their relevance. Leave unavailable metrics blank. Mark new topics as needs_review.

After you review the topics, a separate writing task can work from the same file:

Read topics marked approved in keywords.csv. Prepare one article draft for a topic that has not been drafted. Save it through the connected CMS as a draft, then update the row with its draft reference and status.

What you get: Research and drafting share a visible record of what has been found, approved, and prepared. You can manage the pipeline by updating the file’s statuses.

2. Keep competitor research current

Outcome: A regular briefing of confirmed changes that could affect your product or positioning.

Create a project file listing competitors and the public pages you want checked. Use web browsing or a suitable research connection, then schedule a recurring review.

Review the public sources in competitors.csv. Record changes to pricing, capabilities, and positioning. Include the source URL and date checked. Separate confirmed changes from observations that need verification. Prepare a briefing covering the findings most relevant to our product.

What you get: A dated research record you can return to, plus a briefing focused on changes that deserve attention.

3. Prepare a weekly customer-feedback review

Outcome: A structured summary of recurring customer concerns, feature requests, and questions for your team.

Place a feedback export in a project. If your customer-feedback service offers a suitable MCP server, you can connect it to retrieve the relevant records instead.

Review this week’s feedback. Group recurring issues and feature requests, preserve references to the original records, and distinguish reported problems from suggestions. Save the findings to the project and prepare a short summary for our product meeting.

What you get: A consistent review that preserves the evidence behind each theme. Your team can inspect the source material before deciding what to prioritize.

4. Review local project documentation on a schedule

Outcome: A prioritized documentation report saved inside your repository.

In Cenmax AI Desktop, create a local-only task, select the repository folder, and allow file writes if you want the report saved automatically.

Review the repository’s setup instructions and user documentation. Identify missing steps, outdated references, and unclear explanations. Save a prioritized report to docs/documentation-review.md, referencing the relevant files. Do not change application code.

What you get: A report waiting in your working folder, ready to use in a later coding session. The task runs while the desktop app is open.

5. Build a reusable research brief

Outcome: A source-backed brief that later writing and planning tasks can use.

Create a project for a market, product category, or technical topic. Add your research questions and select the browsing or MCP tools needed to investigate them.

Research the questions in research-questions.md. Save the findings with source links, dates, and unresolved questions. Clearly distinguish documented facts from your interpretation. Prepare a concise brief for someone who has not read the source material.

What you get: A shared foundation for follow-up work. Another chat can use the brief to prepare an outline, compare options, or develop a proposal without restarting the research process.

6. Prepare approved content for publication

Outcome: A repeatable handoff from reviewed content to your publishing system.

Connect a compatible CMS MCP server and keep a project file identifying which content is ready. Start by saving drafts so you can check formatting and destination settings.

Read publication-queue.csv. For entries marked approved_for_draft, create drafts in the connected CMS. Preserve the supplied title and content, record the resulting draft references, and update the queue. Leave all other entries unchanged.

Once you are satisfied with the process, you can create a publishing task for content you explicitly mark ready, where the connected CMS supports publishing.

What you get: A clearer publication queue with references and statuses that subsequent tasks can read.

Choose where each workflow runs

Use cloud-only execution for work involving cloud projects, shared project files, and connected services. It can continue while your desktop app is closed.

Use local-only execution when the task needs to work within a folder on your computer. These tasks require Cenmax AI Desktop to remain open. Local-only describes where the task executes; it can still call Cenmax AI and the MCP servers you select.

Make the output easy to trust and reuse

A useful task instruction defines more than the activity. It specifies the result and the boundaries of the work.

• Name the output: Specify the file, destination, format, and required fields.
• Preserve evidence: Ask research tasks to include sources, dates, and missing information.
• Record progress: Use clear statuses so later tasks can identify what is ready and what remains.
• Select relevant connections: Give each chat or task the servers needed for its outcome.
• Run it once first: Inspect the output before putting the task on a recurring schedule.
• Make external actions explicit: State whether the task should prepare a draft, update a record, or publish approved content.

A simpler alternative to setting up your own agent

If you have considered OpenClaw or Hermes to automate research, maintain working files, or run recurring tasks, you can now build those workflows directly in Cenmax AI.

For the workflows covered here, Cenmax AI can replace the need to install and maintain a separate agent environment. Your projects, connected tools, working files, and cloud schedules are available through one interface.

Self-managed agent setups can involve technical work before the first useful result. OpenClaw’s documented setup includes runtime requirements, installation, onboarding, and running a Gateway. Keeping that Gateway in the background involves installing and checking a service. Those steps can be unfamiliar if your everyday work is marketing, research, content, or business operations.

Both OpenClaw and Hermes offer guided setup options, and Hermes also offers a desktop installer. The distinction is the work you want to take responsibility for: Cenmax AI gives you a managed web workspace for the connected, scheduled workflows described in this post, without requiring you to operate a separate agent installation.

In Cenmax AI, you can:

• Start in your browser: Create a project and add its working files without installing a local agent for cloud workflows.
• Connect services through the interface: Sign in with OAuth where supported, or enter the service’s API key securely. Cenmax AI discovers and activates its tools after your approval.
• Schedule work through a form: Set the instructions, connections, and frequency without configuring a background service yourself.
• Keep the work visible: Use shared files and task history to inspect outputs and decide what should happen next.

A marketer can maintain a keyword pipeline. A researcher can prepare a recurring briefing. A business owner can build a queue of content drafts. Each can start with the outcome and configure the workflow inside Cenmax AI.

You still authorize connected services and define what a task may do. If that is hard, setting up OpenClaw or such agents would be even harder. For cloud workflows, Cenmax AI handles execution while you focus on the instructions, the output, and the decisions that need your attention without you having to worry.

Start with one recurring outcome. Create a project, add its working files, connect the necessary services, and run the task once. When the result meets your needs, put it on a schedule.