Microsoft Project

Horizun Project MCP

An MCP server for Microsoft Project that needs neither Java nor Microsoft Project — and tells you the truth about what it wrote.

It works as a PMO, not as a software operator. It does not just click through Microsoft Project on your behalf: it plans, controls and reports the way a project management office does, following PMBOK — a WBS that obeys the 100% rule, a cost-loaded baseline, earned value and earned schedule, quantitative schedule risk, formal change control with reasons and approvers, weekly look-ahead planning with PPC, and a period status report with its trend. Project is the engine; the method is the product.

Español

Point any MCP client at a .mpp, a Primavera .xer, or an MSPDI .xml and ask real questions: where the critical path runs, which resources are overbooked, whether the schedule would survive a DCMA audit, what a two-week slip actually does to the finish date. Then write changes back — and know which ones landed, because every one is re-read from the model before it is reported.

dotnet tool install -g HorizunMsProjectMcp

That is the whole install. No JVM. No Microsoft Project. No licence. It reads and schedules on its own.

Connect a client

Full per-client instructions, including the plugin route and uninstall, are in docs/INSTALL.md.

Why this one

There are a handful of Microsoft Project MCP servers. Every one of them requires Java, or a licensed Microsoft Project install, or a paid JDBC driver — and on the machine this was built on, none of them would start. Eight things here are not available anywhere else:

The 27 tools

Session — project_health · project_open · project_save

project_health is a doctor, not a ping. It detects Microsoft Project, tests whether its COM server actually starts, and when it does not, hands back the HRESULT diagnosis and the repair steps. That is not hypothetical: this machine hit CO_E_SERVER_EXEC_FAILURE with a perfectly valid registration, and step one of the repair path it emits is what fixed it.

Reading — project_info · tasks_query · links_query · resources_query · timephased_query

Filtered, paged, field-selectable. Tasks are addressed by their stable uid; the row id is display only, because it shifts the moment a task is inserted and addressing by it edits the wrong task.

Analysis — schedule_analyze · schedule_qa · baseline_compare

Computed server-side, so the agent asks a question instead of pulling two thousand tasks into context. Critical path, float distribution, driving path, day-by-day overallocation, DCMA-14, and full earned value (BCWS/BCWP/ACWP, SPI, CPI, EAC, TCPI) — denominated in cost where the schedule carries costs, in work hours where it carries hours, and weighted by duration where it carries neither, which is most of them. The report says which.

Writing — tasks_write · links_write · resources_write · calendars_write · schedule_update

Batched, typed, verified. Cycles are refused before they are applied, with the offending chain named. Calendars are edited in full: the working week and its hours (half or full Saturdays, shifts), working and non-working exceptions with their hours, the project calendar, and task and resource calendars. Tasks take their type, effort-driven flag, fixed cost, a baseline cost that leaves the baseline dates alone, and a whole outline in one batch; resources take rates, overtime rates, material units and max units Project actually sees. schedule_update op level_resources levels with Microsoft Project's own leveller, optionally within slack only. One thing is not offered: moving a task to another position in the outline. Asking for it gets a refusal that names it — an operation that half-works is worse than one that is absent.

Planning — schedule_recovery · schedule_scenarios · schedule_target · schedule_sequence · schedule_learn · schedule_generate

Reprogramming, measured rather than asserted. schedule_recovery finds what is late, ranks it by how much of the schedule sits behind it, then tries each recovery lever — removing lag on the driving chain, overlapping hand-offs once half the predecessor is done, compressing the longest critical tasks, working full Saturdays, and all of them together — on a throwaway copy and reports the finish date each one genuinely produces, and why any lever was dropped. schedule_scenarios compares alternatives you define — calendar, resource, task and logic changes together — side by side: finish, cost, overallocation, critical tasks, nothing committed. schedule_target tests a date you have been handed and names the work the network does not hold in place. schedule_sequence recovers the logic a schedule is missing by reading the order its own dates already state — the planner laid the work out correctly and never linked it, and that decision is recoverable. schedule_learn mines finished schedules for how long each activity actually takes and what usually comes before it; schedule_generate turns that into a first draft, one task per apartment or floor, sequenced the way the history says the trades follow each other.

Feed schedule_learn a model export alongside the schedules and it measures productivity — what a crew actually got through in a day — by joining quantities to tasks on the shared code. schedule_generate then sizes durations from the quantities of the new project rather than copying a remembered duration, because the rate is what carries between projects and the quantity is what changes. Give one export per schedule, in the same order: rates are measured per project, and quantities totalled across projects would inflate every one of them.

A draft can only be as well sequenced as the schedules it learned from. Where the sources link each activity to itself unit after unit but never to the trades around it, both tools say so and name the number: the library reports how many activities learned a predecessor other than themselves, and the draft reports how many trades it left with nothing scheduled before them.

Project control — schedule_risk, plus PMO aspects of the tools above

• Quantitative schedule risk (schedule_risk): Monte Carlo over the network's own logic, all four link types and lags, with three-point ranges per task or a default spread, counted in working hours of the project calendar (half Saturdays and elapsed lags included) and anchored on the schedule's own finish, so the most-likely run is the schedule. P10/P50/P80/P90 finish, probability of a target and of the baseline, P80 contingency, risk drivers by criticality and sensitivity. Seeded runs repeat exactly.
• Period status report (schedule_analyze aspect status_report): planned vs earned, SPI, CPI, earned-schedule SPI(t), EAC by three methods, TCPI, forecast finish, milestones at risk, top issues, and the trend of every recorded cut-off.
• Cost-loaded schedule (schedule_update op load_budget): a budget by code, the same code BIM links on, spread over the tasks as cost; lines no task carries are reported. timephased_query measure s_curve returns cumulative PV, EV and AC, in money when costs are loaded.
• Integrated change control: a rebaseline needs a reason (and takes approvedBy); every baseline, budget load and committed edit batch is logged with its finish impact (aspect change_log).
• Short-interval planning (aspect lookahead): the next weeks' work with each task's constraints, ready vs constrained, and PPC against last week's recorded commitment (Last Planner).
• WBS discipline: schedule_qa checks the 100% rule (empty summaries, cost or resources on summaries, duplicate WBS codes, single-child summaries); project_export format wbs_dictionary writes the WBS dictionary.

The history (cut-offs, commitments, changes) lives beside the schedule in <name>.hzpmo.json, so it survives whichever tool writes the .mpp next.

Interop — project_export · project_import · bim_link · bim_sync

CSV, JSON, MSPDI, Primavera XER and PMXML, native .mpp, and a shaped Power BI dataset. Imports plan before they write.

What it costs to have loaded

The 27 tools present about 13,500 tokens of schema, in every prompt, for as long as the server is connected. That is the honest price of the surface and it is worth knowing before choosing to carry it. It is also why the surface stays near 25: the largest alternative ships 79 tools, and past a point an agent cannot hold the surface in its head well enough to choose correctly within it.

The two contracts

Nothing is applied until it is verified.

{
  "applied": 12,              // operations that fully succeeded, re-read from the model
  "fieldsVerified": 31,
  "rejected": [{
    "uid": 45, "field": "finish",
    "requested": "2026-09-10", "actual": "2026-09-14",
    "reason": "'finish' is calculated from the task's duration and its predecessors. Change the
               duration or the logic, or set a constraint, rather than writing the date."
  }],
  "impact": {
    "tasksMoved": 312,
    "projectFinishBefore": "2027-03-14", "projectFinishAfter": "2027-03-28",
    "criticalPathChanged": true, "newNegativeFloat": 8
  },
  "verifiedBy": "reread"
}

A dry run is a real simulation. dryRun: true deep-copies the schedule, applies the batch, reschedules it with the critical-path engine, measures the difference, and throws the copy away. The impact numbers are observed, not predicted.

Capability honesty

project_health publishes a capability matrix, and the tools honour it. Two things stay false on the file backend and refuse rather than approximate:

• write_native_mpp — no library can author the binary format. Where Microsoft Project is installed, the save is delegated to it and you get a genuine .mpp; where it is not, you get MSPDI and an explanation.
• level_resources — Microsoft Project's levelling heuristic is unpublished. Any imitation would be a different answer wearing the same name.

Saving over a baseline that already holds data is refused too, unless you ask for it explicitly. A baseline is the record of the original plan that every variance is measured against, and it cannot be recovered from the file afterwards.

Everything else — scheduling, recalculation, dry-run simulation, rescheduling incomplete work, the DCMA Critical Path Test — works everywhere. Who computes the dates is the next section.

Dates: Microsoft Project's own

Where Microsoft Project is installed, Microsoft Project calculates the dates. Recalculating, rescheduling after a write, dry runs, recovery options, target dates and the DCMA Critical Path Test all hand the schedule to Project, let it calculate, and take back the dates it computed. project_health reports it as schedulingEngine: microsoft-project.

Measured, not assumed. On seven real construction schedules — 21 to 5,984 tasks, 8,823 in all counting summaries, with split tasks, work in progress, 9-hour days and Saturday half-days — recalculating through this server reproduced the dates Microsoft Project computes on every task of every schedule, to the minute, and left progress exactly as Project had it. Writes were held to the same standard: a duration change, a percent complete or a new link made through this server and the same edit made by hand in Project produced the same dates and the same progress, task for task, summaries included.

That took more than opening the file in Project. Project's own XML import does not reproduce Project: on a real 126-task schedule, re-opening even Project's own XML export put every task on different dates. These are the causes, and the hand-off to Project handles each:

• Unassigned tasks. A task with no resource still carries a placeholder assignment, and on import Project schedules it against the placeholder resource's calendar instead of the task's. It is left out whenever it carries nothing the task does not already say.
• Split tasks. Work that starts, pauses and resumes keeps its gaps only in the assignment's day-by-day work, so that travels too.
• Work in progress. An in-progress task comes back from an XML import with part of its completed work moved into remaining. So progress is never taken back from that trip; when a write changes it, it changes by Project's own rules — a longer duration keeps the work already done, a new percent starts or finishes the task — and summaries roll it up the way Project does.
• Blank rows. A row inserted in Project and left empty is not a task to Project, but it reads as one dated at the project start, and a summary on a real schedule grew back two years to reach it. It goes over as the blank row it is.

A recalculation of the 4,753-task schedule takes about 20 seconds, most of it inside Project.

If you have Project open, it is used as it is. Microsoft Project is single-instance — any automation, this server's included, lands in the copy you are using — so this server never hides your window, never recalculates or closes your documents, and only ever touches a temporary copy it opened itself, checking before every step that the active document is still that copy. You are left on the document you had active.

Without Microsoft Project

MPXJ reads and writes schedule files but does not schedule them, so there is also a critical-path engine of its own — forward and backward pass, total and free float, all four relation types, lag, constraints, deadlines, actual dates, and each task's working calendar. It is what runs on macOS, Linux, and Windows machines without Project.

⚠️ That engine is not Microsoft Project's scheduler

It reproduces Project exactly on schedules built through this server, not on real imported ones: on four production files it matched Project's own start dates on 100%, 69%, 23% and 23% of tasks. It does not implement task types, effort-driven scheduling, resource-driven dates, manual or split tasks, or elapsed durations. So without Project, an imported schedule is never silently rescheduled — its dates stay Project's until you ask for this engine's explicitly with schedule_update op='recalculate', which warns you first. Reading, analysis, DCMA-14 and earned value run on the dates the file already holds and are unaffected.

Wiring it to any other client

Claude Desktop, Claude Code and Codex are covered in docs/INSTALL.md. For anything else that speaks MCP over stdio — Cursor, VS Code, your own client:

{
  "mcpServers": {
    "horizun-msproject": {
      "command": "horizun-msproject-mcp"
    }
  }
}

Registry name: io.github.HorizunGroup/horizun-msproject-mcp (see .mcp/server.json).

Call project_health first in every session — it tells you which backend you are on and what it can do.

Build and test from source

cd src/HorizunMsProjectMcp && dotnet build && cd ../..
python tools/acceptance-test.py   # 65 checks, all tools end to end
python tools/scheduler-test.py    # 45 checks, critical-path engine correctness
python tools/planning-test.py     # 52 checks, reprogramming and learning
python tools/robustness-test.py   # 34 checks, concurrency and hostile input
python tools/smoke-test.py        # 15 checks, environment and capabilities
python tools/packaging-test.py    # 33 checks, the metadata every client reads
python tools/imported-file-test.py   # 16 checks, building on a file Project saved
python tools/planner-features-test.py  # 32 checks, calendars, costs, outline, scenarios, read-only safety
python tools/project-engine-test.py  # 29 checks, Microsoft Project as the engine (needs Project)

321 checks, driven over real JSON-RPC against the running server, on Windows and on Linux. With Microsoft Project installed the suites run with Project calculating the dates, and project-engine-test.py compares the server against Project doing the same thing by hand — dates, progress and summaries, task by task — and checks that a Project you have open is left exactly as it was. The Linux job is the evidence for the headline claim: it runs on a machine with no JVM and no Microsoft Project.

The acceptance suite builds a construction schedule from nothing and asserts the contracts above: that a dry run commits nothing, that a cycle is refused before it is applied, that a write to an unknown uid is rejected rather than ignored, and that the DCMA Critical Path Test moves the finish date by exactly the delay injected into it.

The scheduler suite is the one that earns trust in the dates. It covers start-to-start, finish-to-finish and start-to-finish logic, positive and negative lag, hard and soft constraints, deadlines producing negative float, calendar exceptions actually pushing the schedule out, and a WBS hierarchy with summary rollup, that an imported schedule is never silently rescheduled, and full round trips through Primavera XER and PMXML and through a real binary .mpp — the last written by Microsoft Project itself, read back by MPXJ, with dates, milestone flags, budget codes and dependencies all intact.

Beyond the suites, the server has been driven through a planner's full working cycle on two production construction schedules — a 5,985-task programme and a 170 MB, 1,937-task one — opening, auditing, baselining, recording progress, measuring earned value and exporting the Power BI dataset, and read against files of up to 7,000 tasks.

The robustness suite is the one that matters for trusting this in a real client: forty writes in flight at once on the same document, hostile paths, absurd page sizes, non-latin names. MPXJ's object model is not thread-safe and an MCP client is free to pipeline calls — unguarded, sixty concurrent writes all failed and left the document unusable. Every document now has its own lock, reentrant because several tools are built on others.

A schedule stays in memory until it is closed, and a 170 MB one costs around 400 MB. Reading many retires the oldest document that has nothing unsaved; when every open document has unsaved work, opening another is refused rather than discarding any of it. project_health reports what is open and what it is costing.

Rebuild the installable package with dotnet pack -c Release.

Formats

Reads .mpp .mpt .mpx MSPDI .xml · Primavera .xer .pmxml · Asta .pp · Planner · GanttProject and more, through MPXJ.

Writes MSPDI .xml (Microsoft Project opens it natively) · .mpx · Primavera .xer and .pmxml · Planner · SDEF · JSON · CSV · Power BI dataset · native .mpp where Microsoft Project is installed.

Layout

src/HorizunMsProjectMcp/
  Diagnostics/   COM detection and the environment doctor
  Backends/      MPXJ I/O, the COM bridge, sessions and fingerprints
  Analysis/      critical-path engine, working calendar, DCMA-14, earned value
  Writes/        the verified-write engine
  Bim/           element matching and the 4D bridge
  Tools/         the 27 MCP tools
tools/
  acceptance-test.py   end-to-end across all tools, 65 checks
  scheduler-test.py    engine correctness, format round trips, safety guards, 45 checks
  planning-test.py     recovery, sequencing, target dates, learning, generation, 52 checks
  robustness-test.py   concurrency, malformed input, resource limits, 34 checks
  smoke-test.py        environment and capability matrix, 15 checks
  packaging-test.py    versions, identifiers and client manifests agree, 33 checks
  imported-file-test.py   a schedule Microsoft Project saved: ids, outline, custom fields, capacity, 16 checks
  planner-features-test.py  calendars, costs, outline, scenarios, S-curve, read-only safety, 32 checks
  project-engine-test.py  Microsoft Project as the engine, against Project by hand, 29 checks

Design rationale and the market benchmark that motivated it: DESIGN-TOOL-SURFACE.md and BENCHMARK-MCP-MSPROJECT.md.

Learn to direct it

Horizun Hub teaches, in Spanish, how to direct AI agents in real construction and design work — and how to review what they deliver before you sign it. The course for this MCP is Opera agentes en Microsoft Project: short classes on real construction work, with a verifiable certificate.

New to agents? Level 1 of Opera agentes de IA desde cero is free with an account.

Licence

MIT.