Connect Claude

Add ddx PMO to Claude — Anthropic’s assistant — so you can ask about your projects without leaving the conversation. Ask “what open risks are on the Marunouchi Tower project?” and Claude looks the answer up in ddx PMO and tells you, with the detail from your registers and documents.

This is called a connector. Claude can look things up, and it can ask ddx PMO to draft a document. It cannot change your registers, your files, or your settings — the boundaries are set out in What it can and cannot do below.

Note

There is a short public summary of the connector — what it is, and the whole tool list — at ddxpmo.com/mcp. This page is the how-to.

Before you start

You need three things.

• The connector switched on for your workspace. It is off by default. If you haven’t already asked for it, contact ddx and ask them to enable the Claude connector for your workspace. It is then switched on project by project, so a project nobody has enabled stays invisible even once the workspace has it.
• Your own ddx PMO login. You connect as yourself. There is no shared workspace credential to hand round, and there is nothing to copy or paste except the address below.
• A Claude plan that allows custom connectors. If you can’t find the option described below in Claude’s settings, that is why. On Team and Enterprise plans an owner usually adds the connector once for the whole organisation, and each person then connects their own ddx PMO account to it.

Adding it in Claude

The exact wording differs a little between the Claude desktop, web, and mobile apps, but the shape is the same:

1. In Claude, open Settings → Connectors.

2. Choose to add a custom connector.

3. Give it a name — ddx PMO — and paste this address:

   https://api.ddxpmo.com/v1/mcp

4. Save it, then choose Connect. Claude sends you to ddx PMO to sign in.

What the sign-in screen asks

You are signing in to ddx PMO exactly as you do for the app, through Clerk. The consent screen then asks you to do two things:

• Choose your workspace. If you belong to more than one, the picker matters — the workspace you choose is the one Claude will be able to read. Pick the wrong one and the connector can see nothing useful.
• Approve the access. ddx PMO asks Claude for your name, your email address, and your workspace membership. That is all it asks for, and it is what it needs to know who you are and which workspace to answer for.

Approve it, and Claude returns to the conversation with the connector connected.

Caution

If you see “That organization is not a ddx workspace”, you picked a personal account or a Clerk organisation that isn’t a ddx PMO workspace. Connect again and choose your ddx PMO workspace on the consent screen.

What the connector can do

There are five things, and every question you ask is some combination of them. You never have to name one yourself — you ask in plain English and Claude picks the tools. They are listed here so you know what is and isn’t possible.

Tool	What it does
list_projects	Lists the projects you can see, with code, name and stage.
search_artefacts	Searches one project’s registers, tracker cases, and the full text of ingested sources.
whats_changed_since	Reports what moved in a project’s registers since a date.
generate_document	Starts generating a new version of a project document.
get_document_status	Checks how that generation is getting on.

You don’t have to name a project by its code — Claude reads the project list first and works out which one you mean.

Listing and searching

Search covers the whole project: every register entry ever recorded, connected ClickUp tasks and GitHub issues, and the full text of ingested sources — the bodies of emails and Slack messages, and the contents of uploaded files.

“List my ddx PMO projects.”

The plainest one, and a good first test. You get back each project you can see, with its code, name, and lifecycle stage. Projects where the connector hasn’t been switched on are not listed.

“What open risks on the Marunouchi Tower project mention the façade contractor?”

Searches the risk register by meaning as well as by keyword, so it finds entries that describe the problem without using your exact words.

“Search the Marunouchi Tower project for what the lease actually says about the rent review date, and quote it.”

Searches the text of the uploaded documents themselves, not just their titles, and comes back with the passage.

“Has anything come in by email or Slack about the handover delay on Marunouchi Tower? Quote the messages.”

Searches ingested message bodies. Useful when you remember a conversation happened but not where.

“What decisions were recorded on Marunouchi Tower about the cladding specification, and what was the reasoning?”

Pulls the decision register entries and the notes extracted around them, so you get the reasoning rather than just the outcome.

Asking what changed

whats_changed_since reads the dates on the record itself rather than searching for words, which is why it answers recap questions that search cannot. It covers five kinds of entry — decisions, risks, issues, action items and meeting minutes — and reports two things: what became current since your date, and what stopped being current, because it was superseded or merged into a duplicate.

Give it a plain calendar date. It can’t be in the future.

“What changed on the Marunouchi Tower risk register since 12 June?”

“Recap what happened on Marunouchi Tower since the last steering meeting on 30 May — decisions only.”

It sorts by when things happened on the project, not by when ddx PMO happened to read them, so a minute uploaded three weeks late still lands on the day the meeting was held. That is the difference between a recap you can take to a steering meeting and one that just reflects your ingestion backlog.

Generating a document

“Generate a status report for Marunouchi Tower covering the last 30 days.”

Claude asks you to confirm before it does this — see below for why. It then starts the generation and comes back with an operation id straight away, because a document can take longer to write than one tool call is allowed to run for. Claude polls with get_document_status and tells you when the new version is ready, with its number and title.

A few things worth knowing:

• It makes a new version. It never overwrites or deletes an existing one. The result lands in the project’s Documents area for review, exactly as if you had generated it in the app.
• The kind has to belong to the project’s methodology. Status report, charter, risk register, decision log and the rest for PMBOK; the ISO 19650 and Agile sets for projects on those methodologies. Ask for a kind outside the project’s methodology and it is refused, with the methodology named.
• The covering window applies to status reports only. Other kinds ignore it.
• It needs permission to write documents in that project, and it draws on your workspace’s LLM budget. If you can’t generate documents in the app, you can’t generate them here either.
• Operation records don’t live forever. They expire about an hour after the generation finishes. The document version itself stays in the workspace for good.

Tip

Claude is doing the reading and the writing; the connector only fetches. If an answer looks thin, ask Claude to search again with different words — it will make another lookup rather than guess.

What it can and cannot do

It cannot change your record. Nothing in the connector can create, edit, delete or approve a register entry, upload or delete a file, send anything, or change a setting. This is a property of what is built, not a policy Claude is asked to respect. If someone emails your project a document containing instructions aimed at an AI, the worst it can do through this connector is produce a poor answer.

The one exception only ever adds. generate_document writes a new version of a project document. It cannot overwrite or delete an existing version, and it cannot touch anything else. It is declared to Claude as an action that writes, which is why Claude asks you before each call instead of running it quietly — that confirmation is deliberate, and it is your gate on both the new version and the spend.

It cannot show you anything you couldn’t already see. Every request carries your own identity. Your permissions and your confidentiality clearance travel with you: a project you can’t open in the app can’t be listed or searched here, and material above your clearance is not returned. What you can see through Claude is exactly what you can see in ddx PMO — no more, and no less.

It cannot act as anyone else. Your connection is yours. A colleague’s Claude gets their access, not yours.

It cannot reach a project nobody has shared. Enablement is per project. A project that isn’t enabled won’t appear in the list, and asking for it by name gets a refusal that says so — never an answer that makes the project look empty.

Some things still aren’t wired up. The connector serves the five tools above. It cannot walk the knowledge graph, read portfolio health, or open the Flow view — those exist in the app but are not exposed to Claude.

If it stops working

• You’re asked to sign in again. Normal. Connections that go unused for 30 days are retired automatically. Reconnect and carry on; nothing is lost.
• Claude says it can’t reach ddx PMO. Most often the connector hasn’t been enabled for your workspace. Check that first, then check the address above is exactly right.
• Claude finds nothing in a project you know has content. Check you chose the right workspace when you connected, and that the connector is switched on for that project — a workspace admin does that under Settings → Modules → “MCP connector”.
• Claude says a project can’t be queried from here. That is the per-project switch, and the message names exactly where to turn it on.
• A document generation fails. Ask Claude to check the status again; the reason comes back with it. The commonest causes are a document kind outside the project’s methodology, and the workspace sitting at its LLM budget ceiling.

Where your information goes

Using the connector sends project information out of ddx PMO and into your Claude account. The results of every lookup — register entries, meeting notes, and passages of ingested emails, messages, and documents — are transmitted to Anthropic to answer your question, and your conversation is then held in your Claude account under your own agreement with Anthropic.

That is worth understanding before you connect, and worth raising with your organisation if your project handles confidential material. It is set out in full in section 8 of the ddx PMO Privacy Policy. A workspace that would rather this didn’t happen at all can leave the connector switched off.

Every call is recorded on your side too — who asked, which tool, and on which project — and connector spend is accounted separately from spend inside the app. Workspace admins can see it under Usage.