This guide explains how to connect PDA Platform to Claude, ChatGPT, Gemini, or any MCP-compatible client. PDA Platform exposes 126 AI tools across 18 modules for project delivery — data loading, risk analysis, compliance validation, GMPP reporting, calibration evaluation, four-tier escalation routing, and a full assurance framework. Implements the nine-layer Verified Autonomy framework (see verified-autonomy-overview.md).
There are two connection methods:
- Local (stdio) — install the package on your machine and point your AI client at it. Fast, private, works offline.
- Remote (SSE) — connect to a hosted instance over the internet. No installation required.
Claude Desktop has native MCP support. This is the fastest way to get started.
pip install pm-mcp-serversIf you need GMPP narrative generation (requires an Anthropic API key):
pip install "agent-task-planning[anthropic]"Open your Claude Desktop config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
Add the PDA Platform server:
{
"mcpServers": {
"pda-platform": {
"command": "pda-platform-server",
"args": [],
"env": {
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}Restart the application. You should see PDA Platform listed in the MCP tools panel with 126 tools available.
Try asking Claude:
"What tools do you have available from PDA Platform?"
Claude.ai supports remote MCP servers over SSE. You can connect to a hosted PDA Platform instance without installing anything.
- In Claude.ai, open a conversation
- Click the Integrations icon (plug icon, bottom of the message box)
- Add a custom MCP server with the URL:
https://pda-platform-i33p.onrender.com/sse
- The 126 tools will appear in the conversation
Click the button below to deploy your own PDA Platform server on Render:
Or deploy manually:
git clone https://github.com/antnewman/pda-platform.git
cd pda-platform
pip install -e packages/pm-data-tools
pip install -e "packages/agent-task-planning[mining,anthropic]"
pip install -e packages/pm-mcp-servers
PORT=8080 pda-platform-remoteThe server will start at http://localhost:8080 with endpoints:
/sse— SSE connection for MCP clients/messages— message endpoint for tool calls/health— health check
Claude Code supports MCP servers natively.
pip install pm-mcp-serversIn your project's .claude/settings.json or your global Claude Code config, add:
{
"mcpServers": {
"pda-platform": {
"command": "pda-platform-server",
"args": [],
"env": {
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}The 126 tools are now available in your Claude Code session.
ChatGPT Desktop added MCP support in 2025. Configuration is similar to Claude Desktop.
pip install pm-mcp-serversOpen ChatGPT Desktop settings and navigate to the MCP servers section. Add a new server:
- Name: PDA Platform
- Command:
pda-platform-server - Arguments: (none)
If ChatGPT Desktop requires a config file, the format is:
{
"mcpServers": {
"pda-platform": {
"command": "pda-platform-server",
"args": []
}
}
}Restart ChatGPT Desktop. The PDA Platform tools should appear in the tools panel.
Google has announced MCP support for Gemini. When available, connect using either:
- Local: Point Gemini at
pda-platform-server(same as Claude/ChatGPT) - Remote: Point Gemini at your SSE endpoint:
https://your-instance.onrender.com/sse
PDA Platform works with any MCP-compatible client.
For local (stdio) clients:
pda-platform-serverFor remote (SSE) clients:
PORT=8080 pda-platform-remoteConnect your client to http://localhost:8080/sse.
Once connected, your AI assistant has access to 126 tools across 18 modules:
| Area | Tools | What it does |
|---|---|---|
| Data | 6 | Load project files (MS Project, P6, Jira, Monday, Asana, Smartsheet, GMPP, NISTA), query tasks, get critical path, analyse dependencies |
| Analysis | 6 | Identify risks, forecast completion (Monte Carlo, EVM), detect outliers, assess health, suggest mitigations, compare baselines |
| Validation | 4 | Validate structure, business rules, NISTA compliance, custom rules |
| NISTA | 5 | Generate GMPP quarterly reports, AI narratives, submit to NISTA API, fetch metadata, validate reports |
| Assurance | 27 | Track assumptions and drift, check artefact currency, analyse review findings, monitor confidence, schedule reviews, log overrides, search lessons learned, measure overhead, run workflows, classify programme complexity, assess gate readiness (P14), generate dashboards and ARMM reports |
| Benefits | 10 | Register benefits (P13), track measurements, detect drift, map dependency networks, cascade impact analysis, forecast realisation, assess maturity |
| Variable | Required | Purpose |
|---|---|---|
ANTHROPIC_API_KEY |
For GMPP narratives and review action extraction | Powers AI-generated content |
PORT |
For remote server only | HTTP port (default: 8080) |
NISTA_CLIENT_ID |
For NISTA API submission | NISTA authentication |
NISTA_CLIENT_SECRET |
For NISTA API submission | NISTA authentication |
"No tools showing up"
- Check the server is running:
pda-platform-servershould start without errors - Check your config file path is correct for your platform
- Restart your AI client after adding the config
"Import error on startup"
- Ensure all packages are installed:
pip install pm-mcp-servers - For narrative features:
pip install "agent-task-planning[anthropic]" - For mining features:
pip install "agent-task-planning[mining]"
"Connection refused" (remote)
- Check the server is running:
curl http://localhost:8080/health - Check your port is not blocked by a firewall
- For Render: check the deploy logs in the Render dashboard
Once connected, try these to verify everything works:
- "What PDA Platform tools do you have available?"
- "Load this project file and show me the critical path"
- "Ingest an assumption: construction materials inflation at 2.5% per annum for project DEMO-001"
- "Run a full assurance workflow for project DEMO-001"
PDA Platform is open source: github.com/antnewman/pda-platform