The Toolchain Behind Every CPP Analysis

The same toolkit that produces every CPP analysis.

Critical Path Partners runs on a proprietary forensic-scheduling engine — 19 specialized analysis modules, an MCP server, and a forensic-correctness audit trail built by a working analyst, not a venture-funded startup. You can connect it from Claude. The deeper work — interpretation, expert testimony, deposable opinions — runs through Dana.

Honesty note: the engine is the easy part. The hard part is reading what it tells you.

19
Analysis Modules
~90k
Lines of Code
734
Tests Passing
13
MCP Tools Exposed
What It Covers

Every standard a forensic delay claim has to survive.

Windows analysis, TIA, collapsed-as-built, DCMA-14, Monte Carlo, Eichleay damages — implemented to AACE Recommended Practice 29R-03 and the SCL Delay & Disruption Protocol. Each module names the standard it implements.

AACE MIP 3.3

Windows Analysis

Observational, dynamic, contemporaneous as-is. Per-window critical-path attribution, concurrent-delay matrix, WOET day classifier, slip velocity.

AACE MIP 3.7 · RP 52R-06

Time Impact Analysis

Prospective fragnet insertion, modeled-additive multi-base. Forward-pass float erosion, cumulative impact tracking, owner-vs-contractor delay quantification.

AACE MIP 3.8

Collapsed As-Built

Modeled, subtractive, single simulation. Independent dual-method validation against windows analysis. The but-for analysis opposing counsel always demands.

DCMA 14-Point

Schedule Health

Industry-standard compliance scoring across baseline and current updates. Logic, lags, hard constraints, BEI, missed tasks, float distribution, critical-path reasonableness.

AACE RP 122R-22

Monte Carlo Risk

Probabilistic finish-date forecast. P10/P50/P80/P90 with linear-interpolation percentiles per AACE convention. Sensitivity tornado, mitigation-vs-baseline scenario comparison.

Forensic Damages

Eichleay / Hudson / Emden

Extended home-office overhead per Eichleay (verified federal precedent: ASBCA 5183, Interstate v. West, J.F. Finnegan). Measured-mile productivity loss. Inefficiency factors per MCAA / Schwartzkopf / Wright / Leonard / Ibbs.

Defense Theories

Counter-Claim Analysis

Ten defense theories tested against any owner delay claim — concurrent delay, pacing, no notice, failure to mitigate, causation gap, float ownership, scope-change disguise, self-caused acceleration, LD cap.

Evidence Triage

Claim Workbench

Takes one folder of mixed evidence (P6 update chain + correspondence) and produces a unified analysis: chain-diff manipulation log, rolling baseline, trust score, slip-to-evidence cross-reference with NER-extracted recurring actors.

Forensic-Correctness Discipline

What separates this from "AI-powered" schedule tools.

Tools that promise to "detect manipulation" or flag "fraud risk" automatically are selling something. Forensic analysis works the other way around — every detector discloses what it can and can't conclude. Built into the engine, enforced by tests:

Built In · Enforced by Tests

  • CORRELATION, not CAUSATION. The engine flags statistical patterns. Whether a pattern means intent, fraud, or a data-entry error — that's an analyst's read, not a tool's verdict. Every threshold detector in the suite carries this disclosure.
  • CPP heuristics are disclosed inline. Every percentage gate (60%, 70%, 80%, 15%) names itself as a CPP heuristic and notes that AACE 29R-03 / SCL Protocol publish no numeric cutoff. Opposing counsel can't attack a threshold that disclosed itself.
  • No fabricated case citations. Every case name (Eichleay, Interstate, Malmaison, Hudson, Finnegan), every AACE RP number, every SCL section, every FRE rule is registry-backed. Audit-verified across the entire suite.
  • Empty-state honesty. Forensic deliverables write no file when there are no findings. Empty reports are weapons in cross-examination — the engine refuses to produce them.
  • Audit-trailed math. Per-party day buckets sum to project drift (no cascade-double-counting). Raw values are emitted alongside rounded values so consumers can verify gates without re-running the detector.
Connect From Claude

The engine is exposed as an MCP server.

Thirteen tools, JSON-safe outputs, deployable. If you use Claude (Desktop or API), you can hit the engine directly — run windows analysis, DCMA health checks, collapsed as-built, or Monte Carlo on your own XER files.

Live

Hosted MCP — Try It Yourself

Connect to the CPP forensic engine from your own Claude. Run windows analysis, DCMA-14 health checks, collapsed as-built, Monte Carlo on your own XER files. No signup, no credit card.

Pick whichever applies to you below — same thirteen tools either way.

If you use Claude in your web browser (claude.ai):

https://mcp.criticalpathpartners.ca/mcp

Click your profile → Settings → Connectors → Add custom connector. Paste the URL above and give it a name (e.g. "CPP Forensic"). Requires a Claude Pro or Max plan.

If you use the Claude desktop app (or Cline, Cursor, or another MCP-config tool):

{ "mcpServers": { "cpp-forensic": { "url": "https://mcp.criticalpathpartners.ca/sse" } } }

In Claude Desktop, go to Settings → Developer → Edit Config. Paste the JSON above into your config file (merge with anything already there). Restart Claude Desktop.

Either way, once connected, ask Claude "What MCP tools are available from cpp-forensic?" — you should see all thirteen.

Note: the engine is the easy part. For deposable opinions, expert testimony, or claims that need to survive cross-examination — hire the analyst (see below).

Self-Host the MCP

Open MCP server source available for inspection. AACE-compliant out of the box; deploy on your own infrastructure if your firm requires air-gapped tools.

cpp-forensic-mcp v1.27.0 13 tools · FastMCP · Python 3.12+ Streamable HTTP + SSE transports 35 / 35 server tests passing
Source-access enquiries: dana@criticalpathpartners.ca

Got a real claim? Talk to the analyst, not the tool.

The engine is open. The analysis that wins arbitrations, survives Daubert, and reads naturally on a deposition transcript — that's the part nobody automates. Fixed-fee preliminary assessment in two weeks. No long-term commitment. No hourly burn.