What Claude can do with iyedy
Once connected, Claude reads your diagrams, follows dependencies, runs the checks and prepares changes. What it can do, task by task, and what it never does.
On this page
Once iyedy is connected, you talk to Claude in your own words and Claude picks what to read. Everything goes through your iyedy account: Claude sees the diagrams you can open, with your roles, never more. The examples below use the Claims Platform; ask the same about your own diagrams.
Find and read your diagrams
Claude lists your workspaces and the diagrams you can read, searches their names and contents, and reads a diagram in full: its elements, relationships, deployment and views. It can read a view the way a person does, in words: its zones, the elements in each, the relationships in reading order and the legend. It can also bring back a view as an image.
“Which of my diagrams mention the claims service? Describe the containers view of Claims Platform.”See what an element touches
Claude follows relationships across every diagram you can read, not only the open one. It answers what breaks if an element fails, what uses it directly, or what it depends on, and it can explain the path that leads from one to the other.
“If the claims database goes down, what is affected across our diagrams? Go three hops deep.”Run the checks
Claude runs iyedy's architecture checks on a diagram (availability, security, coupling, completeness, data, operability, hygiene) and reads the findings already stored, with their status. On a large diagram it asks for the counts first, then for the findings of the rules you care about. It can also explain what a rule looks for and why.
“Run the checks on Claims Platform and explain the high findings first.”Compare versions and states
Claude lists the saved versions of a diagram and says what changed since one of them. It reads the current, transition and target states with their dates and work packages, and what separates two states. For a deployment, it can read the network analysis: address plan, routes, and which flows are encrypted.
“What changed in Claims Platform since the version named Before review? What is left between the current and the target state?”Prepare a change
Claude can work out a change, check it first without storing anything (is it valid, which findings does it add or resolve), then submit it as a proposal. A proposal does not change your diagram: it waits in iyedy, labelled with the Claude app that made it, until someone who can edit the diagram applies it or rejects it. Claude can also list the proposals it made and take one back.
“Propose a cache in front of the claims service, check that it adds no high finding, and explain why in the proposal.”By default, Claude asks for your approval before it stores a proposal or takes one back, unless you chose Always allow for that tool. Applying a proposal is always done in iyedy, by a person. See How proposals reach your team.
What Claude never does
- It never edits, deletes or shares a diagram. There is no tool for that.
- It never reads above the ceiling. Up to Confidential on a personal workspace, Internal in a Team or Enterprise organisation. A diagram above it does not exist for Claude, and a Restricted diagram is never sent.
- It never sees secrets. Values that look like passwords or keys are replaced before they leave iyedy.
- It is not iyedy's Assistant. The Assistant works inside the app, on iyedy's credits. Claude works in your Claude app, under your Claude account.
What this means for you
Use Claude for questions that span several diagrams, for quick analyses while you code, and to draft changes. Keep the decision in iyedy: review the proposal there, with its rows and its findings, before you apply it. To see every tool by name, with its parameters, read the MCP tools reference.
Was this page helpful?Your answer is anonymous and counts only if you accepted measurement.
What went wrong?
Don't write personal data.