P6-MCP
MCP serverCommunity
Works on P6 XER files or a live P6 EPPM instance with one tool set for CPM, DCMA, earned value and baseline diffs. By Alireza Shamshiri.
1 starsUpdated 7 weeks agoGitHub data from 5 Nov 2026
At a glance
- Type
- MCP server
- Works with
- Primavera P6
- How it connects
- Reads and writes files; the tool isn't needed
- Access
- Read-only or read-write (configurable)
- Tool versions
- unknown (XER format and P6 EPPM; no version range stated)
- Operating systems
- Windows, macOS, Linux
- Needs a Primavera P6 licence
- No
- Transport
- Local (stdio), Remote (HTTP)
- AI clients documented
- Claude Desktop, Cursor, Windsurf, VS Code, Zed, Continue, OpenAI Agents SDK
- Licence
- MIT
- Language
- Python
What it can do
- Parses and writes XER files, including calendars, and limits file access to the folders you list in P6MCP_WORKSPACE_DIRS.
- Runs a DCMA 14-point assessment, critical path and float analysis, earned value and resource histograms from either backend.
- Compares two XER snapshots, or an XER against a live P6 EPPM project.
- Against P6 EPPM, can also run schedule and levelling jobs, apply actuals and create baselines.
- Includes a command line (inspect, dcma, diff, evm, export) that works without an AI client.
Limitations
- Write-back tools stay off until you set P6MCP_ENABLE_MUTATION=true, and each change needs confirm=True.
- Scheduling, levelling and applying actuals are only available against P6 EPPM, not XER files.
- The README doesn't state which P6 or EPPM versions were tested.
Install
{
"mcpServers": {
"p6-mcp": {
"command": "uvx",
"args": ["p6-mcp", "serve", "--transport", "stdio"],
"env": {
"P6MCP_WORKSPACE_DIRS": "/path/to/your/xer/files"
}
}
}
}Source: github.com/shamshirialireza/P6-MCP#option-a-claude-desktop-recommended-for-most-users
docker run -p 8000:8000 -v /path/to/xer:/data ghcr.io/shamshirialireza/p6-mcpSource: github.com/shamshirialireza/P6-MCP#option-c-docker-remoteserver-deployments
MCP servers and skills run code on your computer. Review the source before installing.
AI-generated models, drawings and results must be checked by a qualified engineer before use.