Skip to main content

Float MCP [beta] prompt guide

Get insights using conversational AI

Justyna Kawalec avatar
Written by Justyna Kawalec
Updated today

Float MCP is currently in beta. Responses may not always be fully accurate, especially for large or complex requests.

We’d love your feedback! Tell us what works, what doesn’t, and which prompts are most useful for your team.

The Float MCP lets you ask natural-language questions about your team directly from LLM tools like ChatGPT or Claude.

It provides secure, read-only access to your Float data and is designed to help Resource Managers, Operations, and Finance teams get fast insights into your team's operations. 

Learn more in the Float MCP [beta] article.

How it works

Float MCP is built on top of the Float public API. Think of it as a conversational layer that lets LLM query Float data through API endpoints.

  1. You write a question/prompt.

  2. The LLM decides which API endpoint(s) to call.

  3. MCP calls Float’s API.

  4. Data is returned.

  5. The LLM interprets and formats the response

Best practices

Float MCP performs best for:

  • Quick snapshots of availability, utilization, and workload

  • “Who / what / when” questions

  • Summaries of schedules, projects, and teams

  • Light reasoning and comparisons when done step-by-step

  • Combining Float data with other MCP-enabled tools (e.g., Jira, Linear)

Because Float MCP is read-only, phrasing matters. To get the most accurate results:

  • Break large requests into smaller steps - guide the AI through the logic instead of asking one complex question

  • Be explicit - timeframes, teams, roles, filters, output format

  • Cross-check important outputs - validate before making decisions

  • Keep complexity manageable - large datasets + multi-step reasoning reduces accuracy.

Because our MCP is built around the Float public API, the prompt accuracy improves when you understand what’s available. You can review Float’s API documentation to see the available endpoints, fields, and IDs.

📝Note: Float’s API uses IDs as primary references, so the AI may return IDs instead of names. If needed, ask the AI to resolve the IDs so the corresponding names are displayed.

Based on your team’s size, data volume, structural complexity, or how clearly the prompt is written, the same query can perform differently. Larger datasets and multi-step reasoning increase the risk of incomplete or inaccurate results. If a result looks incorrect, reword the prompt and guide the AI more explicitly: clarify filters (e.g., active people only), specify the time range explicitly, or request output in smaller chunks (by team, week, or role).

For example, if the following prompt underperforms:

  • Show everyone with at least one full day free next week.

Try guiding MCP step by step:

  1. Fetch all scheduled time-offs for next week.

  2. Filter people who have at least one full weekday off next week.

Or make it even more structured:

  1. Fetch all people from the Marketing department.

  2. Fetch their time off data for next week.

  3. Filter people who have at least one full weekday off next week.

  4. Repeat for other departments.

This reduces the reasoning load and improves retrieval accuracy.

Limitations

Float MCP is currently in beta and may show limited performance with:

  • Complex, multi-step tasks

  • Large data volumes

  • Company-wide insights in larger organizations

When the LLM is asked to fetch multiple datasets, combine them, and calculate results (especially with filters), there’s a risk of incomplete retrieval, inaccuracies, or hallucinations.

Prompt cookbook

Start with a simple question. If you need more detailed insights, refine your request in steps. Use AI follow-up questions to dig deeper or narrow your focus as you go.

Below are some example prompts you can use to quickly get insights from your Float data using MCP.

Quick checks

Use these for fast answers and daily standups:

  • Who’s on time off next week?

  • What projects am I scheduled for today?

  • Which projects start in the next two weeks?

  • Who has no scheduled work tomorrow? Exclude archived users from availability checks automatically.

  • Show my team’s schedule for today, grouped by person.

Capacity and availability

Great for resourcing and planning work:

  • Show everyone with at least one full day free next week.

  • For next month, list the top five people with the most available hours.

  • Who is scheduled above their capacity tomorrow?

  • Show all team members who are on leave or public holidays this month.

Resource conflicts and gaps

Helpful when managing overload or missing assignments:

  • Which people are overbooked next week?

  • List projects next month that have scheduled work for roles but no actual assigned people.

  • Show any roles where only one person is available next week.

Utilization and trends

Useful for lightweight reviews:

  • Compare utilization this week vs last week for the Marketing team.

  • Which team’s scheduled hours increased the most compared to last month?

  • Show average utilization by department for the past four weeks.

  • Compare the March vs. the February workload.

Project health & budgets

Spot risks early:

  • Which active projects have the highest scheduled hours in the next 30 days?

  • List active projects with no scheduled work next month.

  • Show projects where scheduled work has sharply increased compared to last month.

Reporting (step by step)

For more advanced questions, guide MCP through the logic:

  1. For April, list scheduled hours by person in the Frontend department.

  2. Group those hours by project and show totals.

  3. Flag any project where one person is scheduled for more than 70% of the work.

Role-specific prompt examples

Different teams use Float data in different ways. Below are examples tailored to common roles.

Resource Manager prompts

Capacity and availability

  • For next week, list each person’s scheduled hours vs capacity, sorted highest to lowest.

  • Who has availability for a two-day (16h) task this week, and which days are free?

Leave and coverage

  • In the next six weeks, where does planned time off reduce coverage below one person per role?

  • List all overlapping leave periods in the Design team this quarter.

Conflict resolution

  • Show people scheduled above 90% for three or more consecutive weeks in Q1 2026.

  • Which projects could be moved by one week to reduce over-allocation next month?

Operations / Delivery prompts

Workload and utilization

  • Show week-by-week utilization for each department over the last eight weeks.

  • Which department had the biggest increase in scheduled hours compared to last month?

Project oversight

  • List projects with no scheduled work in the next two weeks.

  • Which projects have the highest number of people assigned this month?

Planning

  • For the next eight weeks, list the top three roles where demand exceeds capacity.

  • If we start a new project on March 15, when do we first have capacity for two designers and one PM?

Finance and Commercial prompts

Utilization and billability

  • For last month, show billable vs non-billable scheduled hours by department.

  • Identify people with unusually low billable utilization this month (bottom 10%).

Budget and margins

  • List active projects with high utilization and long remaining timelines.

  • Rank active projects by budget burn rate based on scheduled hours over the next four weeks.

Combining Float MCP with other tools

If you use other MCP-enabled tools (such as Jira, Linear, or Notion), you can combine insights across systems:

  • Fetch planned projects from Linear and compare them with available capacity in Float.

You can also combine Float data with an external public context. For example:

  • For a Website Redesign project, fetch designers from Float.com and evaluate relevant experience based on public LinkedIn profiles.

Workflow prompts examples

Over-capacity remediation

  1. Identify overages

    For next month, list people scheduled over 100% weekly capacity, grouped by week. Include person, role, scheduled %, and top contributing project.

  2. Pick a person

    For [PERSON], break down next month by week: projects + hours per project. Highlight the biggest driver.

  3. Find alternatives

    Find 3 people with the same (or closest) role and at least [X] free hours in that same week. List free hours and current projects.

  4. Draft a move

    Suggest a reallocation option: move [PROJECT/TASK] hours from [PERSON] to [CANDIDATE] for [DATES]. Summarize before/after capacity for both.

Underutilization cleanup

  1. Identify underused capacity

    For next week, list people scheduled under 60% of weekly capacity, grouped by role. Include person, role, scheduled %, and free hours.

  2. Pick a person

    For [PERSON], break down next week by day: projects, scheduled hours, and free hours.

  3. Find demand

    List projects next week that require [ROLE] support and have people scheduled over 90% capacity.

  4. Draft a move

    Suggest a reallocation option: assign [X] hours from [PROJECT] to [PERSON] on [DATES]. Summarize before/after capacity for both.

Holiday coverage

  1. Identify upcoming leave

    In the next 6 weeks, list all planned time off grouped by week and role.

  2. Spot fragile coverage

    For those weeks, identify roles where only one person is scheduled (excluding people on leave).

  3. Inspect the risk

    For [ROLE] in [WEEK], list who is scheduled, their hours, and remaining capacity.

  4. Draft a mitigation

    Suggest a coverage option for [ROLE] in [WEEK] by reallocating hours or assigning a backup. Summarize impact on capacity.

Project pressure

  1. Identify high-pressure projects

    Which active projects have the highest scheduled hours in the next 30 days? Include project, total hours, and number of people assigned.

  2. Pick a project

    For [PROJECT], break down the next 4 weeks by person: role, hours per week, and % of their capacity.

  3. Find bottlenecks

    Identify people on [PROJECT] scheduled above 85% capacity during this period.

  4. Draft a relief option

    Suggest one adjustment to reduce risk on [PROJECT] (reallocate work, add capacity, or shift timing). Summarize before/after capacity.

Team health pulse

  1. Establish baseline

    Show average utilization for the [DEPARTMENT] this week.

  2. Compare over time

    Show the same metric for last week and two weeks ago.

  3. Identify drivers

    List the top 3 people contributing most to the change this week.

  4. Draft a summary

    Summarize the team’s current health in 3 bullets: pressure points, slack, and risks.

Hiring signal

  1. Identify demand pressure

    For the next 8 weeks, list roles where scheduled demand exceeds available capacity, grouped by role.

  2. Check persistence

    Identify roles that remain over capacity for 3 or more consecutive weeks.

  3. Trace demand sources

    For [ROLE], list the top projects driving demand over the next 8 weeks.

  4. Draft a hiring narrative

    Summarize whether this represents temporary demand or sustained need, based on scheduling patterns.

Low billable utilization

  1. Identify low utilization

    This month, list people with billable utilization below [X]%, grouped by department and roles.

  2. Pick a person

    For [PERSON], show billable vs non-billable scheduled hours for this month.

  3. Compare to peers

    Show average billable utilization for [ROLE] in [DEPARTMENT] this month.

  4. Draft an improvement

    Suggest one way to improve [PERSON]’s billable utilization using existing project demand.

Did this answer your question?