bolnet

Use when the user wants a recorded video walkthrough of the Decision- Optimization Diagnostic (DX) upload UI — drop CSVs, watch the pipeline run, see the report rendered. Drives the local pe-mcp-web app via Playwright with caption overlays at each step, records to .webm, and prints a one-line ffmpeg command to convert to MP4. Reproducible, scripted, no manual screen capture required.

Use when a PE shop or portco needs to audit a deployed AI-agent fleet — inventory every registered agent, flag zombies (idle too long), runaway-cost agents (modeled monthly spend over a threshold), and misaligned agents (eval rubric fail), and produce a board-defendable pruning recommendation list with annual savings if the prunes land. Tackles the 40%-of-agentic- projects-cancelled-by-2027 risk Gartner is forecasting, and the agent-sprawl problem mega-funds (Vista, Thoma Bravo) face once they deploy AI agents at portco scale. Pure deterministic — no LLM call inside the tool, modeled telemetry stamped as modeled.

Use when an operating partner wants to surface cross-portco

Use when an LP has sent the GP an AI-section DDQ (Due-Diligence Questionnaire) — typically the new ILPA v2.0 (Q1 2026) AI governance / data / risk sections — and the GP needs a first-draft response packet built deterministically from the fund's existing AI-evidence artifacts in finance_output/, with every answer citing its source and a cross-answer consistency layer flagging contradictions before the LP does.

Use when a portco is post-close and the operating partner needs

Use when a PE professional needs an EU AI Act (Regulation 2024/1689)

Use when a PE professional needs a fast diligence pass over a

Use when you need to grade a PE document AI output (CIM extractor, DDQ generator, IC memo drafter, board memo) against its structured source-of-truth. Scores four dimensions — citation accuracy, hallucination rate, coverage, consistency — without an LLM call, so the grade is reproducible and defensible. Closes the trust gap that stops PE shops from putting AI output in front of an IC.

Enterprise memory layer for agent teams — durable, deterministic recall with bi-temporal facts, RBAC, and audit trails. Self-hosted Postgres + Pinecone + Neo4j; zero LLM in the critical path.

Use when the user asks for financial analysis, stock data, price charts, returns, volatility, risk metrics (Sharpe ratio, drawdown, beta), multi-ticker comparison, correlation heatmap, CSV data exploration, liquidity risk modeling, investor classification, or environment/package validation. Activate automatically with the /finance command.

Agent memory and enterprise features for ArangoDB Community Edition. Use when storing, searching, or managing agent memories in ArangoDB, or when the user asks about CDC, audit logging, vector search, graph traversals, encryption validation, or backup for ArangoDB.