mpptool

Read Microsoft Project and Primavera P6 files without either program. mpptool is a small, fast command-line tool written in Rust. It can:

• open .mpp files from Project 98 through Project 365, plus .mpx and Microsoft Project XML,
• open Primavera P6 .xer exports and P6 XML files, including files that hold several projects,
• print summaries, tasks, resources, assignments and calendars as tables, JSON or CSV,
• convert any of those formats to Microsoft Project XML (MSPDI), which Microsoft Project opens directly (P6 → MS Project included),
• build a new, scheduled project plan from a simple CSV, with stable activity IDs, holiday and shift calendars (10-hour days, Saturday overtime), total/free float and the critical path,
• recompute CPM (cpm) and run a Monte Carlo schedule-risk analysis (risk: P50/P80/P90, criticality, sensitivity, threats and opportunities with optional cost, joint cost/schedule confidence) on any schedule it can read,
• support lean and agile practice: Last Planner PPC and constraint log (lps), takt plans and flowline charts (takt), WSJF portfolio ranking (wsjf) and throughput forecasting (forecast),
• check a schedule (DCMA-style check), compare versions, report variance, resource loading and boards, all on the schedule's own calendar.

It works on Windows, macOS and Linux, and it is designed so both people and AI agents can script it and improve it.

mpptool cannot write native .mpp files. Microsoft hasn't published the format. Instead it writes Microsoft Project XML, which Project opens with File → Open and can then save as .mpp.

---

Download for Windows

1. Go to the latest release and download mpptool-vX.Y.Z-windows-x86_64.zip.
2. Right-click the zip → Extract All…, for example into C:\Tools\mpptool.
3. Open a terminal in that folder: open the folder in File Explorer, click the address bar, type cmd, and press Enter. Alternatively, Shift + right-click inside the folder and choose Open in Terminal.
4. Run it, using the full path to your file:

mpptool.exe summary "C:\Users\me\Documents\My Plan.mpp"
mpptool.exe tasks "C:\Users\me\Documents\My Plan.mpp"
mpptool.exe tasks --csv "C:\Users\me\Documents\My Plan.mpp" > tasks.csv
mpptool.exe to-xml "C:\Users\me\Documents\My Plan.mpp" "My Plan.xml"
mpptool.exe summary "C:\Users\me\Documents\Schedule.xer"
mpptool.exe to-xml "C:\Users\me\Documents\Schedule.xer" "Schedule for MS Project.xml"

tasks.csv opens in Excel. My Plan.xml opens in Microsoft Project (File → Open, then choose XML Format as the file type).

The binary is not code-signed yet, so Windows SmartScreen may warn you the first time. Click More info → Run anyway. To check the download, compare certutil -hashfile mpptool-...zip SHA256 with SHA256SUMS.txt on the release page.

macOS and Linux builds are on the same release page. Extract the archive and run ./mpptool. On macOS you may need to run xattr -d com.apple.quarantine mpptool first.

Install from source

# Requires Rust 1.88+ (https://rustup.rs)
cargo install --git https://github.com/cruiseliner/mpptool
# or, from a clone:
cargo build --release   # binary: target/release/mpptool

Usage

mpptool summary     <file> [--json]           Properties, counts, start/finish, cost, % complete, critical tasks
mpptool tasks       <file> [--json|--csv]     Every task: ID, UID, outline level, dates, duration, %, flags,
                                              predecessors (by visible ID and by UID), slack, constraint,
                                              deadline, actuals, cost, baseline, notes
mpptool resources   <file> [--json|--csv]     Resources: type, max units, rate, email, group, calendar
mpptool assignments <file> [--json]           Who is assigned to which task: units, work, cost
mpptool calendars   <file> [--json]           Working week and exceptions (holidays) of each calendar
mpptool projects    <file>                    Projects in the file (Primavera files can hold several)
mpptool to-xml      <file> <out.xml> [--pin-dates]   Convert .mpp/.mpx/.xml to Microsoft Project XML
mpptool from-csv    <in.csv> <out.xml>        Schedule a new plan from CSV (calendar options: see below)
mpptool risk        <in.csv> [--json]         Monte Carlo finish dates (P10–P90) and criticality index
mpptool holidays    --country US --years 2026-2028   List public holidays
mpptool gantt       <file> <out.html|out.svg> [--critical-only] [--from D] [--to D] [--title T]
                                              Gantt chart: offline HTML page or plain SVG
mpptool check       <file> [--json] [--status-date D] [--fail-on-error]
                                              DCMA 14-point schedule assessment
mpptool compare     <old> <new> [--format text|json|html] [-o out] [--match auto|uid|activity-id|name]
                                              Diff two versions of a plan
mpptool variance    <file> [--json|--csv] [--status-date D]
                                              Baseline vs current/actual slip per task, earned value
mpptool resources-load <file> <out.html|out.csv>  Weekly hours per resource, over-allocation flagged
mpptool to-xlsx     <file> <out.xlsx>         Excel workbook for Excel / Smartsheet
mpptool board       <file> <out.html> [--json] [--status-date D] [--title T]
                                              Kanban board by status with RAG flags
mpptool portfolio   <folder> <out.html|out.json>   Dashboard of every readable plan in a folder
mpptool lps         <workplan.csv> [--constraints c.csv] [--plan plan] [--json]
                                              Last Planner PPC, reasons for variance, constraint log, make-ready
mpptool takt        <trades.csv> --start D [--takt N] [--mode takt|flow] [--html out] [--plan-csv out]
                                              Takt plan / flowline schedule and chart
mpptool wsjf        <portfolio.csv> [--budget B] [--by-cd3] [--json]   WSJF / CD3 ranking with funding line
mpptool forecast    <history.csv> --remaining N [--by D] [--json]      Throughput Monte Carlo forecast

Global options:
  --locale de|fr|...      Language of .mpx files (en, de, fr, it, pt, sv, zh, es, ru)
  --project <id|name>     Project to read from a multi-project Primavera file (default: the first one)

The file format is detected from the file's content. The .xer extension is used only as a fallback.

Input	Reader
.mpp (Project 98–365), .mpx, Microsoft Project XML	mpxj-rs
Primavera P6 .xer, P6 XML (APIBusinessObjects/PMXML)	mpptool's own reader (src/read_p6.rs)

Examples

$ mpptool tasks plan.xml
   ID   UID LVL Name                                          Start            Finish             Dur(d)    % F  Predecessors (IDs)
    3     3 3       Wireframes                                2026-10-19 08:00 2026-10-21 17:00     3.00    0 C
    4     4 3       Visual design                             2026-10-22 08:00 2026-10-27 17:00     4.00    0 C  3
    7     7 3       Backend                                   2026-10-20 08:00 2026-10-27 17:00     6.00    0    3SS+1d

F shows flags: M = milestone, C = critical. Predecessors use the IDs you see in Project. The CSV and JSON output also include the stable UIDs.

mpptool summary --json plan.mpp | jq .critical_path      # scripting / AI agents
mpptool tasks --csv plan.mpp > tasks.csv                  # open in Excel / pandas
mpptool to-xml plan.mpp plan.xml                          # hand to Microsoft Project, ProjectLibre, ...

Gantt charts

mpptool gantt plan.mpp plan.html                    # self-contained HTML (inline SVG, no external assets, works offline)
mpptool gantt plan.xer plan.svg --critical-only     # plain SVG with only the critical path (and its summary rows)
mpptool gantt plan.mpp q3.html --from 2026-07-01 --to 2026-09-30 --title "Q3 look-ahead"

The chart shows WBS/outline rows, task bars with % complete fill, summary brackets, milestones as diamonds, link arrows (FS/SS/FF/SF), the critical path in red, a dashed "today" line (--today D overrides the date), and a grey baseline ghost bar under each task that has a baseline. The time scale adapts to the span: days (≤ 45 days), weeks (≤ 200 days), months (≤ 900 days) or quarters. Hover a bar for its dates. Very long plans are scaled down to keep the chart under 24,000 px wide; thousands of rows render in well under a second.

Schedule health check (DCMA 14-point)

mpptool check plan.mpp                       # table
mpptool check plan.xer --json                # machine-readable: status, threshold, count, %, offending task IDs
mpptool check plan.mpp --status-date 2026-06-30 --fail-on-error   # for CI: exit 1 if any check fails

#	Check	Threshold	Needs
1	Logic: incomplete tasks missing a predecessor or successor	≤ 5%	–
2	Leads (negative lag)	0	–
3	Lags (positive lag)	≤ 5% of relationships	–
4	Relationship types	≥ 90% FS	–
5	Hard constraints (MSO, MFO, SNLT, FNLT)	≤ 5%	–
6	High float (> 44 working days)	≤ 5%	total float
7	Negative float	0	total float
8	High duration (> 44 working days)	≤ 5%	–
9	Invalid dates: actuals after, or unstarted/unfinished forecasts before, the status date	0	status date
10	Resources: tasks with duration and no assignment	0%	resources in the plan
11	Missed tasks: baselined to finish by the status date but finished (or forecast) late	≤ 5%	baseline + status date
12	Critical path test (continuity, see below)	continuous	critical flags
13	CPLI = (remaining critical path length + finish float) / length	≥ 0.95	total float
14	BEI = completed tasks / tasks baselined to finish by the status date	≥ 0.95	baseline + status date

Population: non-summary tasks; checks 1–8 and 10 use incomplete tasks, relationship checks use links into incomplete tasks. The status date comes from the file (MS Project status date, P6 data date) or --status-date. Checks whose data is missing are reported as skipped with the reason, never as failures. The exit code is always 0 unless --fail-on-error is given. Notes: check 12 is a static approximation — DCMA's real test delays a critical task and re-schedules, which needs a scheduling engine; mpptool instead verifies that the project's finish activity is critical and traces the critical chain back from it. CPLI estimates working days as 5/7 of calendar days.

Comparing versions and baseline variance

mpptool compare plan-v1.mpp plan-v2.mpp                     # text: added/removed tasks, changed fields
mpptool compare old.xer new.xer --format html -o diff.html  # self-contained HTML report
mpptool compare a.mpp b.mpp --format json --match name      # match tasks by name instead of ID
mpptool variance plan.mpp                                   # slip per task vs baseline + earned value
mpptool variance plan.mpp --csv --status-date 2026-06-30 > variance.csv

compare matches tasks by unique ID (Microsoft Project UIDs are stable across saves), or by Activity ID when both files have them (Primavera, or plans from from-csv with an activity_id column). Activity IDs are read from any MS Project custom text field whose alias is "Activity ID" (mpptool writes Text1), so they survive an MS Project re-save even if UIDs are renumbered or the field moves; use --match activity-id to insist on it. It reports tasks added and removed, and for matched tasks changes to name, start, finish (with the shift in working days), duration, % complete, milestone flag and predecessor links (type and lag), plus the movement of the project finish. Every report also says how many matched tasks were keyed by activity ID versus UID (matched 41 by activity_id, 2 by uid (fallback); JSON coverage) and warns when tasks without an activity ID fell back to UID matching, which is how a deleted alias shows up (as a task removed and re-added). It also flags UID-matched pairs whose names differ (case and surrounding spaces ignored), the usual sign of mis-pairing after a re-save renumbered UIDs (JSON coverage.name_mismatches). --strict makes either finding exit with code 1 after the report is written; otherwise they are warnings and the exit code is 0.

variance lists every task with a baseline: baseline vs current dates (actual dates when the task has them), start and finish slip in working days of the project calendar, and duration variance. When the file has baseline costs it adds earned value at the status date (from the file or --status-date):

Metric	Meaning
BAC	baseline cost (budget at completion)
PV (BCWS)	baseline cost × planned fraction at the status date (linear over the baseline's working days)
EV (BCWP)	baseline cost × % complete
AC (ACWP)	actual cost
SV / CV	EV − PV / EV − AC
SPI / CPI	EV / PV / EV / AC
EAC	BAC / CPI (typical)
EAC (atypical)	AC + (BAC − EV)
EAC (CPI×SPI)	AC + (BAC − EV) / (CPI × SPI)
TCPI	(BAC − EV) / (BAC − AC): efficiency needed to finish on budget
ES, SV(t), SPI(t)	earned schedule: the working day at which baseline PV equals EV; ES − AT; ES / AT
IEAC(t)	planned duration / SPI(t), in working days, with a forecast finish date

Totals are summed over non-summary tasks. Primavera files carry no baselines (P6 keeps baselines as separate projects), so variance reports "no baseline" for them; compare an older export with the current one instead.

Resource loading

mpptool resources-load plan.mpp load.html   # heat map: resources × weeks, red = above weekly capacity
mpptool resources-load plan.xer load.csv    # resource_uid,resource,week_start,hours,capacity_hours,over_allocated

Each assignment's work is spread over the working hours of the project calendar and summed per week (weeks start on Monday). Weekly capacity is the resource's max units × that week's calendar hours (40 h on the Standard calendar, less in holiday weeks). Material and cost resources are left out. Over-allocation is judged per week, not per day.

Excel and Smartsheet export

mpptool to-xlsx plan.mpp plan.xlsx

The workbook has a Tasks sheet and a Resources sheet. Tasks columns: Row, Task Name (indented by outline level, summary rows in bold), Outline Level, WBS, Activity ID, Start, Finish, Duration (days), % Complete, Predecessors, Assigned To, Milestone, Critical, Summary, Baseline Start, Baseline Finish, ID, Notes. Predecessors refer to row numbers of the sheet (row 1 = first task), Smartsheet-style: 3, 5SS, 7FS +2d, 4FF -1d. In Excel, summary rows are also collapsible row groups, the header row is frozen and has filters.

Importing into Smartsheet

1. In Smartsheet choose Import → Microsoft Excel, pick plan.xlsx, sheet Tasks, and use row 1 as column headers. Make Task Name the primary column.
2. Open Project Settings, turn on dependencies and map Start Date → Start, End Date → Finish, Duration → Duration (days), Predecessors → Predecessors, % Complete → % Complete. Because rows keep their order, the row-number predecessors line up with Smartsheet's row numbers.
3. Smartsheet's Excel import does not rebuild parent/child rows from indentation. Re-create the hierarchy by selecting the rows of each phase and choosing Indent (sort or filter by Outline Level to find them), or, if you need the hierarchy and links rebuilt automatically, import Microsoft Project XML instead: mpptool to-xml plan.mpp plan.xml, then Import → Microsoft Project in Smartsheet.
4. With dependencies on, Smartsheet re-calculates dates from durations and predecessors using its own calendar; set the working days and holidays in Project Settings first so dates match.

Kanban board

mpptool board plan.mpp board.html                       # Not started / In progress / Done
mpptool board plan.xer --json --status-date 2026-06-30  # the cards as JSON

Each non-summary task becomes a card in Not started (0%), In progress or Done (100% or an actual finish), with its parent summary, dates, % complete, resources and a RAG flag judged at the status date (from the file, --status-date, or today):

• Red: unfinished and past its finish date, finishing more than 5 working days later than its baseline, or negative float.
• Amber: not started although its start date has passed, 1–5 working days late against the baseline, or a critical task in progress.
• Green: everything else, including all done tasks.

Portfolio dashboard

mpptool portfolio ./plans portfolio.html    # every .mpp/.mpx/.xml/.xer below ./plans
mpptool portfolio ./plans portfolio.json

One row per plan (per project for multi-project Primavera files): start, finish, % complete, finish slip against the baseline (working days), critical task count, task count and a health score (the % of the applicable DCMA 14-point checks that pass, as in check). A plan is red when it slips more than 10 working days or scores below 50, amber when it slips at all or scores below 75. Files that are not readable plans (other XML, password-protected files) are listed separately rather than stopping the run.

Primavera P6

mpptool projects schedule.xer                       # list projects in the file
mpptool --project BRIDGE tasks schedule.xer         # pick one by id, short name or name
mpptool --project BRIDGE to-xml schedule.xer bridge.xml   # open bridge.xml in Microsoft Project

How P6 data is mapped:

P6	mpptool / Microsoft Project
WBS (PROJWBS / WBS)	Summary tasks, with outline level and dotted WBS code. Each WBS node's activities come first, then its child WBS nodes
Activity (TASK / Activity)	Task: name, start/finish (actual, else remaining/early, else planned), at-completion duration, actual dates, total float
Activity ID (task_code / Id)	activity_id field. In MS Project XML it goes to Text1, with the alias "Activity ID"
% complete	According to the activity's type: duration ((orig − remaining) / orig), physical, or units
Start / Finish milestones	Milestones (zero duration)
Critical	Not complete and total float ≤ 0
Primary constraint	Start on → MSO, Start on or after → SNET, Start on or before → SNLT, Finish on → MFO, Finish on or after → FNET, Finish on or before → FNLT, ALAP, Mandatory start/finish → MSO/MFO
Relationships PR_FS/SS/FF/SF	FS/SS/FF/SF links with lag (hours; shown in days when whole)
Calendars (clndr_data / work week + holidays)	Calendars with working times and exceptions. P6 hh:59 end times are rounded to the next whole minute
Resources, RSRCRATE / ResourceRate	Resources (labor/non-labor → Work, material → Material) with their earliest rate
Resource assignments	Assignments with units/time, work and cost. Task cost = sum of assignment costs

Creating a plan from CSV

name,start,duration_days,predecessors,outline_level,resources,notes
Website Launch,,,,1,,
Wireframes,2026-10-19,3,,2,Alice,Low-fidelity first
Visual design,,4,2,2,Alice;Bob,
Backend,,6,2SS+1,2,Carol,
Go live,,0,3;4,2,,

mpptool from-csv plan.csv plan.xml

• activity_id (or id, optional): a stable ID such as A1010. IDs must be unique and are written to the XML as Text1 ("Activity ID"). Use them instead of row numbers so inserting a row doesn't break links, progress updates or risk links.
• predecessors: activity IDs (A1010, A1010SS+2) or row numbers starting at 1 (#3 forces a row number when IDs are numeric), separated by ;. Each can have a link type (FS is the default, or SS, FF, SF) and a lag in working days, e.g. 2SS+1 or 4FF-1d.
• start (YYYY-MM-DD, optional): saved as a Start No Earlier Than constraint. The earliest start becomes the project start. If no row has a start, the project starts today.
• outline_level: a row followed by deeper rows becomes a summary task.
• duration_days: 0 makes a milestone.
• resources: created automatically, each at 100% units.
• notes: optional free text.
• optimistic, pessimistic (days, optional): three-point estimates used by risk.
• By default scheduling uses Microsoft Project's Standard calendar (Mon–Fri, 08:00–12:00 and 13:00–17:00, no holidays), so existing plans schedule exactly as before. mpptool computes total and free slack, writes both to the XML (TotalSlack, FreeSlack) and flags critical tasks.

Calendars: holidays, weekend work and overtime shifts

from-csv and risk take the same calendar options. Dates are rescheduled around all non-working time, and the calendar (weekly pattern plus every holiday/exception in the plan's years) is written into the MSPDI XML, so Microsoft Project shows the same calendar.

Option	Meaning
--holidays US	Skip public holidays. US = US federal holidays (5 U.S.C. 6103) with the observed-date rule (Saturday → Friday, Sunday → Monday). Also CA, GB, AU, DE, FR, IN, MX; combine with commas (US,CA). Default none.
--years 2026-2028	Only these years' holidays (default: every year the schedule spans).
--holiday-file days.csv	Custom exceptions, one per line: date[..date],name[,hours]. hours empty/off = non-working; 07:00-15:00 (or 06:00-10:00;10:30-16:30) = working with those hours; work = a normal working day. Custom days override holidays and the weekly pattern.
--work-days mon-sat	Working weekdays: mon-fri (default), mon-sat, all, mon,tue,thu, …
--hours 06:00-11:00,11:30-16:30	Daily working hours (default 08:00-12:00,13:00-17:00).
--day sat=07:00-12:00	Override days of the weekly pattern (repeatable): Saturday overtime, mon-thu=06:00-16:30 (4×10s), fri=off.
--hours-per-day 8	Working hours in one duration_days day (default 8). Longer shifts finish work faster: a 3-day (24 h) task takes 2.4 ten-hour days.
--calendar-name	Calendar name in the XML (default Standard, or Project Calendar when customised).

# US federal holidays, 10-hour weekday shifts plus Saturday overtime, a year-end shutdown
mpptool from-csv plan.csv plan.xml --holidays US --years 2026-2028 \
  --hours 06:00-11:00,11:30-16:30 --day sat=07:00-12:00 --holiday-file shutdown.csv

mpptool holidays --country US --years 2026-2028      # list them (also --csv, --json, --country list)

# shutdown.csv
2026-12-28..2026-12-31,Year-end shutdown
2026-11-28,Overtime Saturday,07:00-15:00

Built-in holiday tables cover national/federal holidays only. Regional holidays (US state, Canadian province, German Land, Australian state), one-off proclaimed holidays and India's lunar-calendar holidays are not included: add them with --holiday-file.

Critical path and schedule risk (Monte Carlo) on any schedule

mpptool cpm  schedule.mpp [--json]                    # forward/backward pass: early dates, total & free float
mpptool risk schedule.xer --runs 10000 --seed 1 [--opt 0.8 --pess 1.5] [--json]
mpptool risk plan.csv --holidays US                   # CSV plans take the calendar options above
mpptool tasks --json schedule.mpp > tasks.json && mpptool risk tasks.json   # simple JSON task list as input

Both commands work on every format mpptool reads (.mpp, .mpx, MSPDI .xml, Primavera .xer and P6 XML), on from-csv plans, and on a JSON task list (the tasks --json output, or the same shape: uid, name, outline_level, duration_hours, predecessors [predecessor_uid, link_type, lag_hours], optional start, summary, percent_complete, constraint_type, constraint_date). Imported schedules are re-scheduled on their own project calendar (working week and exceptions); --holidays/--holiday-file add to it. Durations are remaining durations; finished tasks are anchored at their finish and in-progress tasks at the status date. This is mpptool's own CPM, not Project's or P6's scheduling engine: task/resource calendars, ALAP and finish constraints, resource leveling and elapsed durations are not modelled, so dates can differ from the source tool. Use it for float and risk analysis, not as a replacement for F9 in Project or F9/Schedule in P6.

Each run samples every task's duration from a PERT-beta distribution (optimistic, most likely = duration_days, pessimistic) and reschedules the network with the same links, lags, start constraints and calendar as from-csv. Optimistic/pessimistic come from the CSV columns, or default to 0.8× and 1.5× the duration. The report gives the deterministic finish, the probability of meeting it, the mean and P10/P50/P80/P90 finish dates, and each task's criticality index: the share of runs in which it was on the driving path to the finish. Tasks that are not critical in the deterministic plan but have a high criticality index are hidden schedule risk. The random generator is built in, so the same seed gives the same result on every platform.

Uncertainty classes, risk events and sensitivity

Add an uncertainty column to the CSV (or optimistic/pessimistic columns, which win) so each task gets a distribution that matches its kind of uncertainty. Built-in classes, as (optimistic, pessimistic) multiples of the most likely duration. These are planning defaults, not industry statistics, so calibrate them on your own history and override with --class vendor=0.95,2.5:

Class	O	P	Use for
labor	0.85	1.4	crew work driven by production rates
equipment	0.9	1.5	equipment-driven work
design	0.8	1.6	design, engineering, reviews
vendor	0.95	2.0	vendor-quoted lead times, fabrication, delivery
permit	0.9	2.0	permits, AHJ reviews, work by others
weather	0.9	1.8	weather-exposed exterior work
fixed	1.0	1.0	concrete cure, notice periods

Tasks without a class use --opt/--pess. Discrete risks from the RAID log go in a CSV via --risks risks.csv:

id,name,probability,min_days,likely_days,max_days,tasks,cost_min,cost_likely,cost_max
R1,Switchgear ships late,30%,5,10,20,A1050,0,5000,15000
R2,Inspector backlog (common cause),0.25,2,3,6,15;16,,,

The cost columns are optional. With them, risk reports each event's expected cost, the risk-event cost distribution (mean/P50/P80/P90) and the joint P80: the share of runs that make both the P80 date and the P80 cost. Only the event costs are counted. Duration-driven cost (overtime premiums, extended general conditions, productivity loss on long shifts) is not modelled, so add it yourself.

In each run, an event fires with its probability. When it fires, one impact is sampled and added to every linked task (activity IDs or row numbers; a summary row means all of its tasks), so one cause hitting several tasks is modelled as one correlated hit. The report adds each task's sensitivity (Pearson correlation of its sampled duration with the finish, shown as a tornado list) and, for each event, its hit rate and mean finish impact (mean finish when it fires minus when it does not).

Opportunities, threat vs opportunity effect, and joint cost/schedule confidence (v0.11)

Risk events can be opportunities (time or cost saved). Add a type column (threat or opportunity) or write the impact as negative days. Both rows below mean "saves 1 to 3 days, most likely 2":

id,name,type,probability,min_days,likely_days,max_days,tasks,cost_min,cost_likely,cost_max
O1,Prefabricated rough-in assemblies,opportunity,50%,1,2,3,A1110,0,3000,6000
O2,Early AHJ pre-inspection walk,,40%,-3,-2,-1,A1500,,,

When an opportunity fires, its sampled saving shortens every linked task. A task never goes below 0 days, and tasks in the fixed uncertainty class (concrete cure, notice periods) are never shortened. The report lists them as "fixed task(s) not shortened". Opportunity cost columns are savings: positive numbers are read as savings, and negative numbers mean the same thing. Impacts that straddle zero are rejected, so split them into a threat and an opportunity.

risk then reports threats and opportunities in separate sections, each with probability, hit rate, mean finish impact and sensitivity (correlation of the event's sampled impact with the finish: positive for threats that drive it, negative for opportunities that pull it in). An event effect block schedules every run's sampled durations three times: with no events, with threats only, and with all events. This uses common random numbers, so the threat effect (+x wd at P80), the opportunity effect (−y wd at P80) and the net effect aren't sampling noise. JSON: events[].type, events[].sensitivity, events[].mean_sampled_days, events[].fixed_tasks_protected, event_effect (uncertainty_only, with_threats, with_all_events, threat_p80_days, opportunity_p80_days, net_p80_days, mean and cost effects). A risk file with threats only gives the same results as v0.10.1 for the same seed.

Joint cost and schedule confidence level (JCL). --base-cost (the point estimate) and --cost-per-day (a time-dependent burn rate, such as general conditions, field overhead and standby) give each run a total cost: base + rate × (finish − deterministic finish, in working days) + the costs of the events that fired. This is a simple form of the integrated cost and schedule risk analysis described in AACE RP 57R-09. The report gives the total-cost P50/P70/P80/P90, the joint confidence at P50/P70/P80, and the JCL of meeting --target-date and --target-cost together (default: the P80 date and P80 cost). It also shows each target's probability on its own. NASA policy (NPR 7120.5F) budgets major projects at a 70% JCL. Uncertainty in quantities and unit rates in the base estimate is not sampled, so this JCL is narrower than a full cost-risk model.

mpptool risk plan.csv --holidays US --risks raid.csv --base-cost 2400000 --cost-per-day 6500 \
    --target-date 2027-02-05 --target-cost 2450000

Uncertainty for imported schedules

.mpp/.xer/XML files carry no three-point estimates or classes, so risk on them uses --opt/--pess. Give a sidecar CSV keyed by activity ID (or row, or name) to keep your uncertainty after handoff:

activity_id,uncertainty,optimistic,pessimistic
A1050,vendor,,
A1080,,4,9

mpptool risk schedule.mpp --uncertainty uncertainty.csv --risks risks.csv

The report states the sidecar's coverage, so a stale sidecar (tasks added in MS Project, rows for deleted tasks) is visible rather than silent:

Uncertainty sidecar: 2 task(s) covered, 1 fell back to --opt/--pess, 1 sidecar row(s) matched no task
  WARNING: sidecar does not cover 1 task(s): A30 Commission
  WARNING: 1 sidecar row(s) match no task (stale?): A99

--json carries the same data in uncertainty_coverage (tasks_covered, fallback_tasks, unmatched_rows). Rows that match no task are reported, not fatal, and the JSON lists are complete (the text report shows 10 of each).

Key rules: activity IDs and names match after trimming whitespace, case-insensitively, and nothing looser. Under an activity_id header a key is only an activity ID (use a row column or #N for row numbers). A key that looks like a plan ID with lost leading zeros (A50 for A0050) or a prefix stripped by Excel (50) is reported as did you mean A0050? (not applied). Two sidecar rows that hit the same task are an error that lists the lines. An activity ID used by several tasks in the plan is a warning, and a sidecar row keyed by it is an error (key by row).

--strict exits with code 1 after printing the report when any task fell back to --opt/--pess or any sidecar row matched no task. Use it in scripts to stop on a stale sidecar.

Status updates (progress.csv)

mpptool from-csv plan.csv update.xml --progress progress.csv --status-date 2026-10-23
mpptool cpm  plan.csv --progress progress.csv --status-date 2026-10-23
mpptool risk plan.csv --progress progress.csv --status-date 2026-10-23

id,actual_start,actual_finish,remaining_days,percent_complete
3,2026-10-19,2026-10-21,,100
4,2026-10-22,,,50
7,2026-10-20,,4,

id (or activity_id) is the task's activity ID when the plan has one, otherwise its row number (or use a name column). Finished tasks keep their actual dates. In-progress tasks keep their actual start, and their remaining work (remaining_days, else duration × (1 − % complete)) is scheduled after the status date. Unstarted work never starts before the status date. The XML carries actual start/finish, % complete and the status date, so check tests 9/11/14 and variance work on the update. Imported schedules (.mpp/.xer/…) bring their own actuals and status date.

Resource leveling (within float)

from-csv --level (and cpm --level) delays tasks whose named resources are double-booked, in early-start order, but only within their total float, so the project finish never moves. Each resource is one unit, busy with one task at a time. Leveled tasks get an SNET constraint at their new start and a [leveled by mpptool] note. Conflicts with no float left are listed, not resolved: add crew, resequence, or accept a later finish. Because leveling adds SNET constraints, keep an unleveled copy of the plan for cpm/risk (constraints hide float), and say which file you hand over. This is a simple serial leveler, not Project's or P6's leveling engine (no partial units, splitting or priorities).

Calendars in reports

compare, variance, board, portfolio and the CPLI in check count working days on the schedule's own project calendar (holidays, 6-day weeks), not plain Mon–Fri. resources-load measures each week's capacity from the calendar: a 5×10 h + Saturday 5 h calendar gives 55 h per resource, and holiday weeks have less.

DCMA options

check --exclude-endpoints stops check 1 counting the project start milestone (no predecessors) and finish milestone (no successors). check --thresholds logic=10,lags=10,fs=85,hard=5,high_float=60,high_float_pct=5,high_duration=30,high_duration_pct=5,missed=5,cpli=0.9,bei=0.9 overrides any threshold. The defaults are DCMA's.

Last Planner System® metrics (lps)

Weekly work plan (week, activity, responsible, committed, completed, reason) and an optional constraint log (id, activity_id, description, category, owner, raised, need_by, resolved):

mpptool lps workplan.csv --constraints constraints.csv --status-date 2026-11-27
mpptool lps workplan.csv --constraints constraints.csv --plan plan.csv --holidays US --lookahead 6 --status-date 2026-11-27

The report gives PPC (Percent Plan Complete: assignments completed as promised ÷ assignments committed, as defined by the Lean Construction Institute) by week, overall, for the last 4 weeks and as a week-over-week trend. It also gives PPC by responsible party (lowest first) and a reasons-for-variance Pareto (free text is grouped into the usual categories: prerequisite work, design/RFI, submittal, materials, labor, equipment, weather, inspection/permit, space, owner change, underestimated effort). From the constraint log it reports open, overdue and resolved constraints, the share resolved by their need-by date, the mean days to resolve, open constraints by owner and by category, and the overdue items. With --plan (any schedule mpptool reads), activities whose early start falls in the lookahead window are checked against open constraints (matched by activity ID or exact name), giving the share made ready and a list of blocked activities. Rows with committed = no are left out of PPC. Missed commitments with no reason are flagged.

Takt plans and flowline (location-based) schedules (takt)

mpptool takt trades.csv --start 2027-01-04 --holidays US --takt 3 --html flowline.html --plan-csv takt-plan.csv
mpptool takt trades.csv --start 2027-01-04 --mode flow --buffer 1          # continuous crews (line of balance)
mpptool from-csv takt-plan.csv takt.xml --holidays US                         # into MS Project

Input is trade,zone,days (trades in sequence order and zones in flow order, as they first appear), or trade,days with --zones 6 or --zones "L1 east,L1 west,L2". In takt mode, every trade moves through the zones at one rhythm: duration = (trades + buffer wagons + zones − 1) × takt. Trade-zone cells whose work exceeds the takt are flagged as overloaded, which means the zone or crew needs rebalancing. Train load is total work ÷ the capacity of the takt grid. In flow mode, each trade keeps one continuous crew at its own pace and starts as early as it can without overtaking the previous trade in any zone (plus --buffer days). The HTML page draws the flowline chart, with zones on the y axis and working days and dates on the x axis. --plan-csv writes a from-csv plan with activity IDs T<trade>-Z<zone>, links and SNET starts.

Portfolio and backlog prioritization (wsjf)

mpptool wsjf portfolio.csv --budget 1700000
mpptool wsjf portfolio.csv --by-cd3 --json

Columns: id, name, business_value, time_criticality, risk_reduction, job_size (relative scores on the SAFe modified Fibonacci scale, so scores off the scale are noted), with optional cost, cod_per_week and duration_weeks. WSJF = (BV + TC + RR/OE) ÷ job size. Ties go to the smaller job. With cost and --budget, items are funded in rank order and the cut line is shown. With cod_per_week and duration_weeks (Reinertsen's economic cost of delay), --by-cd3 ranks by CD3 = CoD ÷ duration, and the report compares the total cost of delay of the ranked order against the input order.

Throughput forecasting (forecast)

mpptool forecast closed-per-week.csv --remaining 60 --start 2026-10-19 --split 1,1.2 --by 2026-12-31
mpptool forecast finished-dates.csv --period day --holidays US --remaining 40

Monte Carlo "when will it be done?" from throughput history, for any countable work: stories, punch-list items, submittals reviewed, RFIs closed. Each run resamples past periods' completions until the remaining count, multiplied by a sampled scope-growth/split factor, is done. The output gives the dates at 50/70/85/95% confidence, and with --by it also gives how many items will be done by that date at each confidence level. History is a completed column per period, or one finished date per item (binned, with empty periods counted as zero). --last N uses recent history only. Fewer than 10 samples gives a warning.

What carries over in to-xml

Carried over	Not carried over (yet)
Tasks, outline, WBS, start/finish/duration, % complete, milestones, summaries, manual scheduling, blank rows	Custom fields / extended attributes
Dependencies (FS/SS/FF/SF, lag)	Baselines 1–10 (baseline 0 is kept)
Constraints, deadlines, actual start/finish	Timephased data, splits, recurring task definitions
Notes (plain text; rich formatting is lost)	Views, tables, filters, formatting
Costs: task cost, fixed cost, actual cost; resource rate, cost per use; assignment cost	Overtime rates, cost rate tables B–E
Calendars: base and resource calendars, working times, exceptions (recurring exceptions are expanded into individual dates)	Work weeks (alternate working periods)
Resources (work/material/cost), assignments (units, work)	Resource availability tables, leveling data

Before release, every one of 606 sample .mpp files from MPXJ's test suite converted to well-formed XML that reads back with identical tasks, links, constraints, deadlines, actuals, notes and calendar exceptions. All 12 Primavera samples in MPXJ's suite (7 XER, 5 P6 XML) also convert and read back identically. The XML output has not yet been verified by opening it in real Microsoft Project (neither the converted files nor the calendars, holidays and float written by from-csv). Before importing, read the XML back with mpptool tasks out.xml (and mpptool calendars out.xml) to confirm tasks, links, dates and calendar exceptions. Please report any file that Project rejects.

Primavera P6: mpptool reads XER and P6 XML but has no XER writer yet. To get a plan into P6, import the Microsoft Project XML that mpptool writes (P6's import of MSP XML varies by version, so verify the result on your P6 version).

Limitations

• No native .mpp output; only MSPDI XML.
• Password-protected .mpp files cannot be read.
• Asta, Phoenix, Planner and other formats are not supported.
• Primavera: these are not mapped: secondary constraints, activity codes, UDFs, steps, expenses, baselines (P6 baselines live in separate projects), resource curves, and level-of-effort / WBS-summary semantics (both are read as ordinary tasks). Durations are at-completion hours, not recalculated against calendars. Relationships to activities in other projects are dropped. XER from P6 versions older than 6 may use an older calendar format that is not parsed; those calendars fall back to Mon–Fri 08:00–17:00.
• The from-csv scheduler is intentionally simple. It uses one project calendar (no task or resource calendars), has no resource leveling, and only supports SNET constraints. Overtime is modelled as extra working time on the calendar; separate overtime work and overtime cost rates are not written.
• cpm/risk on imported files use only the project calendar and SNET/MSO constraints (see above).
• risk samples task durations independently (risk events give common-cause correlation; there is no general correlation matrix). The JCL cost model is base + time-dependent rate + event costs; base-estimate uncertainty is not sampled.
• lps, takt, wsjf and forecast are planning aids over CSV data; they don't replace pull-planning sessions, a takt production system, a portfolio governance process or a team's own flow metrics tool.
• Duration months are converted as 20 working days.
• check test 12 (critical path test) is a static continuity check, not DCMA's delay-and-reschedule test.
• compare/variance count slip on the project calendar (no task calendars); PV and earned schedule assume work spread evenly over each baseline, not the timephased baseline.
• resources-load spreads work evenly (no work contours, timephased data or per-resource calendars).
• board and portfolio RAG rules are fixed (documented above), not configurable yet.
• The readers are young (2026) ports of MPXJ, so expect rough edges on unusual files.

Roadmap

• Check XML in Microsoft Project automatically, or against a corrected MSPDI schema
• Custom fields, baselines 1–10, work weeks, timephased data in to-xml
• --format json export of the whole project in one document
• Gantt charts (gantt) and DCMA 14-point check (check) — v0.3.0
• Diff two schedules (compare) and baseline variance / earned value (variance) — v0.4.0
• Resource loading (resources-load) and Excel/Smartsheet export (to-xlsx) — v0.5.0
• Kanban board (board) and portfolio dashboard (portfolio) — v0.6.0
• Holiday calendars, weekend/overtime work patterns, float in XML, Monte Carlo risk — v0.7.0
• Risk events, uncertainty classes, sensitivity, progress updates, leveling within float, DCMA options — v0.8.0
• Opportunities, threat/opportunity effect, JCL with time-dependent cost; Last Planner (lps), takt/flowline (takt), WSJF (wsjf), throughput forecast (forecast) — v0.11.0
• XER writer (next: TASK/TASKPRED/CALENDAR/PROJWBS tables)
• Earned value with timephased baselines; EVMS (EIA-748-E) guideline checks
• Cost-estimate uncertainty (quantity × unit-rate ranges) in the JCL; correlation between task durations
• 4D/VDC export: activity-ID ↔ model-element link table (IFC GUIDs) for Navisworks/Synchro
• Reference-class uplift (risk --uplift) from a user-supplied reference class
• Portfolio dashboards that include PPC, JCL and benefits/OKR fields
• Holiday-aware compare/variance slip; regional holiday tables; correlation between task durations
• Primavera: activity codes/UDFs as custom fields, baselines, expenses, cross-project links, writing XER
• Code-signed Windows binaries; winget / Homebrew packages

Contributions are welcome. See CONTRIBUTING.md. AI coding agents should read AGENTS.md.

License

mpptool's own code is dual-licensed under MIT OR Apache-2.0, at your option. See LICENSE-MIT and LICENSE-APACHE.

mpptool depends on mpxj-rs, which is licensed under LGPL-2.1-or-later. Binaries built from this repository therefore include LGPL code. See THIRD-PARTY-NOTICES.md for what that means and how to relink. mpptool is an independent project, not affiliated with Microsoft or with the MPXJ project.