User roles: ✔ Users with the MCP access role ✔ Application Administrator ✔ Integration Administrator
Portfolio questions rarely sit in one project. Which projects carry the most risk? Where is governance slipping? What does the steering committee need to see this month? Answering them usually means opening dashboards, filtering lists and copying numbers into a report.
The Portfolio Health toolset gives an AI app such as Claude or Microsoft Copilot a portfolio-level view of Fluid, so it can answer these questions for you. The app can list the projects in a portfolio, programme or project with their RAG status and governance score, rank them by risk, health or governance, pick out projects with weak governance, and produce an executive summary. The toolset is read only and every tool runs as you, so results include only the projects you can see in Fluid. It's aimed at portfolio managers, PMO leads and executives who want answers and reporting rather than updates. To change a project, use the Project Manager toolset.
In this article you will learn what you need before you start, how the app decides which projects a question covers, what each Portfolio Health tool returns, and how to get good results.
Before You Start
Fluid has made MCP available on your site. If your administrator cannot find MCP Tools and Connected Apps in the Administration Console, contact Fluid support.
MCP access is on. An Application Administrator has turned on Enable MCP Access on the MCP Tools and Connected Apps page. See Managing Fluid AI Connectors (MCP) for Administrators.
You have access. You have the MCP access role, or you are an Application or Integration Administrator. While Administrators only (evaluation) is on, only administrators can connect.
Your AI app is connected to Fluid. See Connect AI apps to Fluid with MCP.
You allowed the right permissions when you connected.
Read portfolio health and executive summaries is ticked by default. Every tool in this toolset needs it.
Read your projects, status reports and schedules is ticked by default. Keep it ticked so the app can find a project or programme you name (see below).
Ask FluidAI questions about your projects (uses AI) is ticked by default. It is only needed for the FluidAI tip at the end of this article.
Your administrator can also switch individual tools off for the whole site. If a tool in this article is missing from your app, check the permissions you allowed, then ask your administrator.
Every tool is listed, with its inputs and the permission it needs, in Fluid MCP (Connectors) Tools Definition.
Scope: which projects a question covers
Every tool in this set takes an optional scope:
A portfolio, programme or project. Fluid identifies it by its ID, not its name. Leave it out to cover every project you can see.
Include child projects. On by default, so a portfolio or programme takes in everything below it. Ask for "this project only" to leave the children out.
Reportable only. Off by default. Limits the results to promoted, reportable projects, as your steering committee views do.
How the app finds the portfolio or programme you name
When you name a portfolio, programme or project, the app finds its ID first. It can do this in two ways:
By name, from your own projects. fluid_pm_find_my_projects, from the Project Manager toolset, finds the projects you are allocated to whose name matches.
From the full list. fluid_pf_get_portfolio_projects with no scope lists every project you can see, each with its portfolio, programme and parent. The app picks out the programme or parent you named and uses its ID, or keeps the projects whose portfolio matches the name.
Tip: ask the app to tell you which portfolio, programme or project it used before it answers. If a name matches more than one, or none, it can ask you which you mean.
Portfolio manager tools
Tool | What it does |
|---|---|
fluid_pf_get_portfolio_projects | Lists the projects in scope, each with its RAG status, governance score and result, and its portfolio, programme and parent. The app gets 50 projects at a time and pages through the rest. |
fluid_pf_get_top_projects | Ranks the projects in scope by one of three measures:
Each project comes with its counts of open impacts, risks, issues and high-severity impacts. It returns 10 projects unless you ask for a different number, up to 100. |
fluid_pf_get_low_governance_projects | Projects whose governance result is below Good, lowest governance score first, with the number of governance groups each one fails or is Borderline on. Borderline projects are included unless you ask to leave them out. A project with no governance result yet counts as Good, so it isn't listed. It returns up to 50 projects unless you ask for a different number, up to 200. |
fluid_pf_get_executive_summary | An executive summary for the scope:
It includes 10 milestones and 10 impacts unless you ask for a different number, up to 50 of each. |
Limits: if you ask for more than a tool allows, for example the top 150 projects, Fluid doesn't refuse. It returns the default number instead, so you would get 10. Ask within the limits, and check how many results came back.
Example prompts
"Find the Digital Transformation portfolio, then give me its executive summary: RAG spread, the next five milestones and the top five open risks."
"Which ten projects carry the most risk right now?"
"Which projects in the Infrastructure programme have poor governance, including Borderline ones? How many governance groups does each one fail?"
"Which milestones in the Digital Transformation portfolio are overdue?"
"List all reportable projects with their RAG and governance score, as a table."
Tips for good results
Ask for the format you need. For example: "as a table", "as three bullet points for the steering committee" or "as a short email to the sponsors".
Combine toolsets. With the Project Manager toolset connected too, the app can go from "which projects carry the most risk?" to "show me the open risks on the top one" in the same conversation.
Use FluidAI for judgement questions. Questions like "why is this portfolio slipping?" are often better answered by FluidAI's portfolio assistants, which you can reach through the Project Manager toolset's fluid_pm_ai_ask_projects tool. It needs the Ask FluidAI questions about your projects (uses AI) permission, which is ticked by default when you connect, and FluidAI must be switched on for your site. Each question uses your site's AI allowance and can take up to a minute, and there's a per-person limit on questions per minute.
Good to Know
Read only. No tool in this set changes anything in Fluid.
Only what you can see. Every result is limited to the projects you can see in Fluid, so two people asking the same question may get different answers.
No financial data. None of these tools returns costs or budgets.
Project search and governance metric tools. Earlier versions of this article listed project search, project summary and governance metric tools. AI apps can't use those. To find projects, the app uses fluid_pm_find_my_projects or fluid_pf_get_portfolio_projects, as described above.