Connect your AI agents

Run your product map from your favorite AI agent.

VPM speaks MCP, the open standard AI agents use to work with other tools. Connect an agent, sign in with your VPM account, and ask it to add a feature, plan a release or tell you what changed.

The address of the VPM MCP server

https://visualproductmanager.com/api/mcp

Your agent learns VPM first

You do not have to explain VPM to your agent. Once connected, it calls vpm_guide, a built-in guide that tells it what VPM is, what features, requirements, modules, releases and versions are, which rules it must respect, and how to work: analyse a product and tell you what is missing, or fill VPM from your description or documents, with a plan for you to approve before it writes anything. Agents that list prompts also offer the guide as vpm_getting_started. After that, just ask: “Analyse my product”, or “Fill VPM from this document”.

Example conversation. You: Which features does the Webshop have? The agent calls list_items (feature · Webshop). Agent: The Webshop has four features: Customer Account, Order Tracking, Returns & Refunds and Support Center. You: Add a feature Gift cards and link it to the Checkout module. The agent calls create_feature (Gift cards · module Checkout). Agent: Done. Gift cards is now a feature of the Webshop, linked to Checkout. It has no requirements yet. Shall I add a few?

What an agent can do

Features and requirements
Add, rename, change the status of or remove them, and link them to modules.
Modules, releases and versions
Plan a release, link what it delivers, record product and module versions, set dependencies.
Specifications
Link the documents behind a feature or requirement, when your organisation uses them.
Show it
A timeline of your releases, a product map and the matrices of the product cube: interactive views in agents that can show apps, text in the others.
Learn VPM first
A built-in guide teaches the agent what VPM is and how to work with it, so you do not have to explain it.
Read everything
Lists, details, links (direct and inherited) and the warnings VPM shows, so the agent can reason about your product first.

Your role decides

An agent acts as you and never has more access than you have. An administrator’s agent works on every product of the organisation. A product manager’s agent only sees and changes the products an administrator assigned to that person. If an administrator changes your role or your products, your agent follows on its next request, with no new sign-in. Nothing ever crosses into another organisation.

Connect in four steps

  1. Add the address above to your agent as a remote MCP server (see your agent below).
  2. Your browser opens the VPM sign-in. Log in with the account you use for VPM.
  3. Check which agent is asking and choose Approve. Decline if you did not start this.
  4. Back in your agent, ask for something: “List my products” is a good first question.

No VPM account yet? Sign up first and join or create your organisation. Any agent that supports remote MCP servers with sign-in will work.

Claude Code

claude mcp add --transport http vpm https://visualproductmanager.com/api/mcp

Then run /mcp inside Claude Code and choose vpm to sign in.

Claude (web and desktop)

Settings, Connectors, Add custom connector. Paste the address above and sign in when asked.

VS Code

{
  "servers": {
    "vpm": { "type": "http", "url": "https://visualproductmanager.com/api/mcp" }
  }
}

Put this in .vscode/mcp.json, then start the server from the editor and sign in.

Cursor

{
  "mcpServers": {
    "vpm": { "url": "https://visualproductmanager.com/api/mcp" }
  }
}

Put this in .cursor/mcp.json (or ~/.cursor/mcp.json), then connect and sign in.

Revoke access

Every agent you approved is listed under My account, in the account menu of the app. Revoke one and it is refused on its very next request. You can connect it again whenever you like.

Good to know

  • An agent can change and remove things, so approve only agents you start yourself and check what it is about to do.
  • Removed items are kept as history, like in the app.
  • Every request an agent makes is recorded, without the content you work on.
  • Requests are limited per user, so a runaway agent cannot overload your workspace.

Everything the server provides

Your agent picks the right function itself; you never have to name one. This is the full list, so you know what it can reach. Every function runs as you, within your role and products.

Learn VPM

How an agent gets to know the concept before it touches your product.

vpm_guide
The guide: what VPM is, the building blocks and their rules, how to analyse a product, how to fill it, and which function does what. The server tells every agent to read it first.
vpm_getting_startedprompt
The same guide as a ready-made start of a conversation, for agents that list prompts (a slash command in many of them).
vpm://guideresource
The guide as a document you can attach to a conversation yourself.

Find your way

Who the agent acts as and which products it can work on.

whoami
Shows your email address, organisation, role and, for a product manager, the products you may work on.
list_products
Lists the products you can access.
get_product
Shows one product: its description, the products it may refer to and how many features, requirements, modules and releases it has.

Read

Everything the app shows in its lists and forms, so the agent can reason about your product before it changes anything.

list_items
Lists the objects of one kind in a product, as the app lists show them: title, status, links and warnings.
get_item
Shows one object with its fields, its links (direct and inherited, with their source) and its release warnings.
list_references
Lists what an object can be linked to, for example the modules a feature may link to.

See

Read-only visuals. In agents that can show apps (for example Claude on the web and desktop) they appear as interactive views you can click through; other agents get the same content as text.

show_timeline
The roadmap: dated releases on a time axis with slip, dependencies and warnings, and the releases without a date in lanes.
show_product_map
The product map: features with their requirements next to the modules. Select an item to see what it connects to.
show_product_views
One matrix: Module coverage (features by modules), Release coverage (features by releases) or Module versions (modules by releases with their versions).

Add and change

Each kind of object has its own pair of functions with the same rules as the app. A change someone else made in the meantime is reported, not overwritten.

create_featureupdate_feature
A feature: what the product does for its users, public or private.
create_requirementupdate_requirement
A requirement of a feature, with its status.
create_moduleupdate_module
A module: a part of how the product is built.
create_releaseupdate_release
A release: target date, status, what it delivers, dependencies, its product version and, when it versions modules independently, a version per module.
create_specificationupdate_specification
A link to the document behind features or requirements. Only when your organisation uses specifications.

Tags

Your organisation's own labels, shared by all its products. Tagging never changes the object itself.

list_tags
Lists the tags with their description, colour and on how many objects they are used.
create_tagupdate_tag
A tag: a name that is unique in your organisation, a description and a colour.
tag_item
Adds tags to a feature, requirement, module, release or specification, or removes them.
delete_tag
Removes a tag from the organisation and from every object that carries it. The agent can first ask on how many objects it is used.

Remove

Removing is soft, as in the app: the item is kept as history.

delete_item
Removes an object together with what hangs under it. The agent can first ask what would be removed, before anything is.

For administrators

Only offered to an administrator's agent.

get_organisationadministrators
Shows the details of your organisation.
list_usersadministrators
Lists the members and open invitations with their role and products.

Not possible, by design: creating, changing or removing products, users, roles or the organisation itself. Those stay in the app.