When to Use
- You already use an AI assistant connected to Cloud Humans via MCP and want to review suggestions without opening the panel.
- You want to approve multiple suggestions at once, chatting instead of clicking on each card.
- You are editing content in the knowledge base and want to know if there are any pending improvement suggestions for it.
Prerequisites
- Improvement Center v2 enabled for your project. The release is done project by project by the Cloud Humans team.
- An AI assistant connected to your account via Cloud Humans MCP.
How It Works
What the assistant can do
The MCP exposes the Improvement Center as tools, so the assistant can:
- See the overview — current retention, projection, and how many suggestions are pending on each front.
- List content suggestions — Gaps, Quality, structural, and duplicates, with the current text alongside the proposed one.
- Fill in the marked sections — suggestions that come with
[COMPLETAR]can be filled in directly within the chat, and the assistant shows the options it found in your own sources. - Approve or dismiss — approving publishes the content to the knowledge base; dismissing removes it from the queue and accepts a note with the reason.
- Decide on agent suggestions — list the suggested Specialist Agents, with the integrations they depend on, and approve or dismiss.
The content page itself tells you
When you open a knowledge base entry, an indicator appears if there is a pending improvement suggestion for it. Previously, you had to go to the Improvement Center to find out.
Notes
- Approving via the assistant has exactly the same effect as approving through the interface: the content enters the knowledge base and ClaudIA starts using it.
- The origin of the decision is recorded, so you can later tell apart what was decided by the assistant and what was decided in the panel.
- A suggestion with an open
[COMPLETAR]is not published. Fill in those sections before approving. - To understand how to prioritize what to review first, see Improvement Center: what the impact number of each suggestion means.