For AI agents

Put agents to work on the enterprise model

Caplo's hosted MCP server lets Cursor, Claude, and your own agents read the repository and propose changes inside a draft. People review and publish. Version history and the audit trail show what changed.

  • Connect any MCP client to one workspace
  • Read the live model, or a draft, from the same repository your architects use
  • Write only into a draft. Publishing stays with a person
  • Every change is on the record: history, snapshots, and an audit trail

How this maps to Gartner's five AI capabilities

How agents work with Caplo

Agents read the model your architects keep, write into a draft, and leave a record of what they changed.

Core benefit

Read the model architects already maintain

Agents do not need a custom export. They can search the repository, follow relations, look at the neighborhood of an entity, and open BPMN. Before they write, they can read the property shapes for this workspace, so new fields match how the team already models.

  • Answer impact questions from the live repository, or from a draft.
  • Follow relations and neighborhoods instead of guessing dependencies.
  • Load process diagrams and the property definitions this workspace uses.

Example view

Grounded in the repository

Ask what depends on an application. The answer comes from the model.

What depends on the back-office suite?

Claims handling

Process

Policy admin

Application

Billing

Application

Relations, neighborhoods, and BPMN. Same repository the architects maintain.

An agent asks what depends on an application and gets the neighborhood from the repository.

Core benefit

Change the model without touching live

Repository writes go into a draft. The agent creates one or reuses an open draft, updates entities, relations, and BPMN there, and returns a link for review. Publish and delete stay in Caplo. The live model changes only when a person decides it should.

  • Per-item history shows what changed and who changed it. You can restore a previous version.
  • Workspace snapshots keep the model as it was at a point in time.
  • An audit trail records creates, updates, deletes, and restores. Agent writes are tagged to the draft they belong to, the same way as an edit in Caplo.

Example view

Review before live

The proposal stays in a draft, and the record starts there.

Draft · Retire back-office suite

  • Updated Application · phase out
  • Added relation · serves
  • Updated BPMN · claims handling

History

What changed, and who changed it

Snapshot

The workspace as it was on a given date

Audit

Create, update, delete, and restore. Tagged to the draft.

Publish stays in Caplo. Restore is there if you need the previous version.

A draft holds the proposal. History, snapshots, and the audit trail record it the same way as an edit in Caplo.

Core benefit

Automate the work that used to wait in a queue

These are workflows you can run with your own agent. Caplo gives it the repository, a draft to write in, and a link for review. Someone still has to publish.

  • Demand management: take a request, map it onto capabilities and applications, and open a draft for review.
  • Repository maintenance: find missing relations, duplicates, and stale descriptions, and propose the fixes in a draft.
  • Scenarios and processes: open a named draft for a retirement or target state, or update a BPMN diagram and return the review link.

Example view

Work you can hand to an agent

The same setup, used on different queues.

Demand intake

Map a request onto capabilities and applications

Repository upkeep

Propose missing relations and stale descriptions

Target-state scenario

A named draft, including a BPMN update

Each one ends as a draft a person can review and publish.

One draft for a demand, a cleanup, or a target-state scenario, ready for someone to review.

Connect

Point an agent at your workspace

One connection, one workspace. The agent works against that repository and hands changes back as a draft.

  1. 1Add https://app.caplo.ai/mcp in Cursor, Claude, or another MCP client.
  2. 2Sign in to Caplo, review the client, and choose the workspace this connection should use.
  3. 3The agent can read that workspace and write into a draft. Review and publish stay in Caplo.

MCP connection guide

https://app.caplo.ai/mcp