project-mcp

project-mcp

MCP server that enables LLMs to read and analyze Microsoft Project schedules, including critical path, resources, and advanced construction planning layers (AWP and LPS) for work packages and Lean planning.

Category
访问服务器

README

project-mcp

CI License: MIT Python 3.11+

🇺🇸 English · 🇧🇷 Versão em português

Model Context Protocol (MCP) server that exposes Microsoft Project files to LLM clients like Claude Desktop and Claude Code. Reads schedules, resources, dependencies, critical-path data, and baseline variance — and adds AWP (Advanced Work Packaging, CII) and LPS (Last Planner System, Lean) layers for work-package planning, constraints, weekly commitments and PPC tracking. All local, no cloud calls, no Microsoft Project license required.

Why

Construction, engineering, and BIM workflows live in Microsoft Project schedules. This server lets your LLM:

  • Inspect a project schedule and answer questions about it (deadlines, critical path, resource overload).
  • Cross-reference task data with quantities from BIM models or with cost data from Power BI dashboards.
  • Generate JSON exports for downstream automation (dashboards, reports, ETL pipelines).

It's read-only by design. Project edits stay where they belong: in Microsoft Project itself.

Requirements

  • Python 3.11+
  • An MCP-compatible client (Claude Desktop, Claude Code, etc.)
  • For .xml (MSPDI): no extra dependencies
  • For .mpp (native Microsoft Project): the optional [mpp] extra (requires a JVM via the mpxj package)

Installation

Option A — uv (recommended)

git clone https://github.com/jeffmodeler/project-mcp.git
cd project-mcp
uv sync

Option B — pip

pip install git+https://github.com/jeffmodeler/project-mcp.git

For .mpp support:

uv sync --extra mpp
# or
pip install "project-mcp[mpp] @ git+https://github.com/jeffmodeler/project-mcp.git"

Claude Desktop integration

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "project-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "C:\\path\\to\\project-mcp",
        "run",
        "project-mcp"
      ]
    }
  }
}

Restart Claude Desktop. The 35 tools become available in any conversation (13 core MS Project + 10 AWP + 12 LPS).

Tools

Tool Purpose
load_project Load an MSPDI .xml or .mpp file into memory
project_info Title, author, schedule window, currency, aggregate counts
list_tasks Filter tasks by type, criticality, name substring, top N
get_task Full record of a single task by UID, ID, or name
list_resources Resources, optionally filtered by type or overallocation
get_resource_assignments Assignments for one or all resources
find_overallocated_resources Resources flagged as overallocated
get_critical_path Tasks on the critical path, sorted by start date
get_predecessors_successors Dependency network for a task
get_baseline_variance Current vs baseline date and duration comparison
get_gantt_data Tasks formatted for Gantt-chart libraries
export_to_json Full project export to JSON file (or inline)
generate_pbip_dashboard Generate a Power BI Project (.pbip) and open it in Power BI Desktop

Exporting .mpp to MSPDI XML

If you don't want to install Java for the optional mpp extra, export your schedule from Microsoft Project as XML:

  1. Open the .mpp file in Microsoft Project.
  2. File → Save As → Save as Type → XML Format (*.xml).
  3. Point load_project at the resulting .xml.

The XML format is the official Microsoft Project Data Interchange (MSPDI) schema and contains tasks, resources, assignments, predecessors, baseline, and most of the project metadata.

Example prompts

After loading a project, ask Claude:

Load the project at C:\schedules\obra-acme.xml
Give me the critical path with total duration in days.
Which resources are overallocated and by how much?
List the 5 tasks with the largest baseline variance.
Export the full project to C:\reports\obra-acme.json

AWP — Advanced Work Packaging

Construction Industry Institute (CII) methodology. Breaks execution into aligned packages across engineering, procurement and field:

CWA (Construction Work Area) → CWP (Construction Work Package)
                                     ↓
                               IWP (Installation Work Package)

Focus: path of construction + readiness (nothing starts in the field until materials, documents and access are available).

AWP tools

Tool Purpose
awp_list_cwa List Construction Work Areas
awp_upsert_cwa Create or update a CWA
awp_list_cwp List CWPs with task_count, total_hours, any_critical
awp_upsert_cwp Create or update a CWP (status: planned/ready/in-progress/complete/on-hold)
awp_assign_task_to_cwp Link a task UID to a CWP (moves it if already elsewhere)
awp_set_cwp_requirements Set CWP requirements (materials, documents, access)
awp_readiness_check Check whether a CWP is ready — compares requirements vs available
awp_path_of_construction Sequence CWPs by earliest start, with critical-task counts
awp_generate_iwps Split a CWP into IWPs sized by labor hours
awp_export_wpr Generate a Work Package Release — self-contained JSON for field teams

LPS — Last Planner System

Lean Construction method with five planning levels:

Master → Phase (pull plan) → Lookahead (N weeks, clears constraints)
                              → WWP (Weekly Work Plan) → Daily huddle

Key metric: PPC (Percent Plan Complete) — commitments kept / commitments made.

LPS tools

Tool Purpose
lps_list_phases List project phases
lps_upsert_phase Create or update a phase (with start/end dates)
lps_set_pull_plan Set execution sequence (pull planning) with task UIDs
lps_get_pull_plan Retrieve a phase's pull plan
lps_register_constraint Register a constraint (material/document/labor/equipment/access/permit/…)
lps_clear_constraint Mark a constraint as resolved
lps_list_constraints List with filters by task, status, type
lps_lookahead N-week horizon with ready/blocked tasks (from open constraints)
lps_add_commitment Add a commitment to a weekly work plan (ISO week e.g. 2026-W03)
lps_mark_complete Close a commitment with actual_hours or variance_reason
lps_get_wwp Read a weekly work plan
lps_ppc Compute PPC for a single week or a series of the last N weeks

Constraint types: material, document, information, design, labor, equipment, access, permit, prerequisite, other.

Variance reasons: weather, design_change, material_delay, labor_unavailable, equipment_breakdown, rework, permit, prerequisite_incomplete, scope_change, other.

Sidecar storage

The .mpp/.xml file remains authoritative (read-only preserved). Next to the project file, an <name>.awp/ folder holds metadata that Microsoft Project does not represent well:

C:\schedules\
├── obra-acme.mpp              ← authoritative schedule (never modified)
└── obra-acme.awp/             ← sidecar folder, created on demand
    ├── awp.json               ← CWA / CWP / IWP hierarchy
    └── lps.json               ← phases, pull plans, constraints, WWPs

Every write updates updated_at (ISO 8601 UTC) in the JSON.

Development

uv sync --extra dev
uv run pytest -v
uv run ruff check src tests

License

MIT — see LICENSE.

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选