Meet Jean-Pierre & Michelle.
AI That Actually Works.¶

Two purpose-built agents that automate project reporting and analytics intelligence.
They connect to your tools, learn your domain, validate every answer, and get measurably smarter every week.
100% on your machine. Zero cloud dependency. Free pilot.

Start Your Free Pilot See How It Works

25+

hrs/week saved per team

3 sec

from question to verified answer

9

anti-hallucination layers

$0

pilot cost — measurable ROI first

What They Automate For You¶

🎩 Jean-Pierre

Project Reporting Automation

Connects to GitHub, Jira, and Slack — turns chaotic project data into a real-time command center with proactive risk scoring and 24/7 background monitoring.

✅ Status reports generated automatically — no manual gathering

✅ Risks detected 2-3 weeks early via real-time scoring (0-100)

✅ Executive reports in one click — CTO / CFO / PMO variants

✅ All projects ranked by risk across your portfolio

✅ Evolution Memory — learns your preferences, never re-explain

~10 hrs/week saved per project manager

Explore Jean-Pierre →

🔬 Michelle

Analytics Intelligence

Connects to your databases — anyone can ask data questions in plain English with verified, source-cited answers backed by a 9-layer anti-hallucination architecture.

✅ Plain English → verified results in 3 seconds

✅ Exact SQL + source tables shown — full auditability

✅ Learns from corrections — never makes the same mistake twice

✅ Shared Brain — knowledge retained, governed via PR review

✅ KPI engine with cron schedules, thresholds, sparklines

~15 hrs/week freed per data team

Explore Michelle →

See the Results¶

🎩 Jean-Pierre

Living Dashboard

Fleet View

Sprint Forge

Delivery Pulse

Team Pulse

Strategic Board

Project Reporting Automation

From chaotic project data → real-time command center with proactive risk scoring, fleet intelligence, and executive-ready reports.

📊Living Dashboard — 24 bento cards, always current

🚨Risk Scoring — 0-100 scale with 2-3 week early warning

📋Fleet View — all projects ranked by risk

📝One-Click Reports — CTO, CFO, PMO variants

🧠Evolution Memory — learns your style, never re-explain

🔬 Michelle

Analytics Dashboard

KPI Dashboard

AI Chat

Schema Browser

Evolutionary Memory

Test Harness

Analytics Intelligence

From ad-hoc SQL requests → self-service analytics with verified answers, interactive dashboards, and automated KPI tracking.

💬Plain English → verified SQL in 3 seconds

📈9 Chart Types — pins, gauges, sparklines, drill-downs

🎯KPI Engine — cron schedules, thresholds, automated alerts

🔍Source Citations — clickable provenance on every number

🧠Shared Brain — team knowledge retained, governed via PRs

Why They Don't Hallucinate¶

🧬

They Evolve

Generic AI resets every session. These agents get measurably smarter every week.

6-namespace Evolution Memory — errors, corrections, tool wisdom, user adaptation, performance, knowledge

Same question, better answer next week — tracked and verifiable

Shared Brain: GitHub-backed, PR-reviewed → when your analyst leaves, knowledge stays

✓

They Validate

Every answer comes with proof — exact SQL, source tables, execution evidence.

9-layer anti-hallucination architecture: Triage → Generate → Execute → Learn

Clickable [source:N] provenance badges on every data answer

Business glossary + rules enforced architecturally — not just in prompts

🔒

They Stay Private

100% on your machine. Nothing — not data, not queries, not memory — leaves.

Single binary, no Docker, no cloud — runs locally on macOS, Linux, Windows

Supports Ollama for fully local AI — zero API calls possible

GDPR-native, air-gap ready, no telemetry, no tracking

9 Layers of Anti-Hallucination¶

Text-to-SQL tools generate queries. Michelle generates queries, validates them, learns from corrections, shares knowledge across your team, and gets measurably more accurate every day.

🧠

Triage

Can it be answered from context?

→

⚡

Generate

SQL with full schema context

→

▶️

Execute

Run with provenance tracking

→

📈

Learn

Record into evolution memory

Adaptive Schema — Tiered injection handles 100+ tables

Business Glossary — Your definitions, injected per prompt

Business Rules — "Exclude deleted" enforced architecturally

Validated Examples — Approved NL→SQL as few-shot training

Evolution Memory — 6 namespaces that persist across sessions

Semantic Joins — LLM-assisted FK discovery beyond constraints

Provenance Tags — Clickable [source:N] citations on every answer

Test Harness — Validate accuracy before deploying to team

Multi-Model Routing — Right model per task: SQL, Q&A, enrichment

"But I Already Have Claude…"¶

Claude is brilliant at conversation. But conversation isn't operations. Here's what changes when you move from a generic AI chatbot to purpose-built operational agents.

Capability	Claude / ChatGPT + MCP	Jean-Pierre & Michelle
Schema knowledge	⚠️ MCP can fetch — no tiered caching	✅ Auto-synced with adaptive tiered injection
Persistent memory	❌ Session resets — SKILLS.md is static	✅ 6-namespace Evolution Memory
Query validation	❌ You validate manually	✅ Triage → Generate → Execute → Learn
KPI tracking	❌ No persistence, schedules, thresholds	✅ KPI engine with cron, history, alerts
Team knowledge	❌ Each user starts from zero	✅ GitHub-backed Shared Brain + PR review
Data sovereignty	❌ Data goes to Anthropic / OpenAI	✅ Local binary, air-gap capable
Interactive dashboards	❌ Text + static artifacts	✅ Charts, pins, KPI gauges, drill-downs
Proactive monitoring	❌ Reactive — you ask, it answers	✅ Background daemon with risk scoring
Business glossary	⚠️ Static SKILLS.md — no enforcement	✅ Dynamic glossary + rules, enforced per query
Improves over time	❌ Same accuracy week 1 and week 52	✅ Measurably smarter every week

The Bottom Line

Claude is a brilliant conversationalist. Jean-Pierre and Michelle are purpose-built operational systems — they live on your machine, learn your domain, validate every answer, and get smarter every week. One is a tool. The other is a teammate.

The Time You're Losing Today¶

Every week, your team loses 25+ hours to manual reporting, ad-hoc data requests, and knowledge gaps. Here's where the time goes.

PMs: 45 min of chaos every morning¶

Open Jira → Slack → GitHub → Spreadsheet. Context-switching just to answer "where do things stand?" — and the data is already stale.

Data teams: buried in lookups¶

15+ ad-hoc SQL requests/week, each 30-60 min. That's 1-2 analysts doing glorified lookups instead of real analysis.

Knowledge walks out the door¶

Your best analyst leaves → metric definitions, query patterns, business context: all gone. New hires spend 3 weeks just learning what "active customer" means.

AI chatbots hallucinate numbers¶

You tried ChatGPT / Claude for analytics. It invents data, forgets corrections next session, and every team member starts from zero. The issue isn't AI. It's stateless AI.

The Time You Get Back¶

Side-by-side: what your team does manually today vs. what the agents automate.

Before

~3 hrs/week

Status gathering

After

10 sec

Automated

Before

~4 hrs/week

Meeting prep + reports

After

One click

Instant generation

Before

30-60 min each, 15/week

Ad-hoc data requests

After

3 sec

Self-service

Before

3 weeks onboarding

New hire learns definitions

After

Instant

Shared Brain has it all

~25 hrs/week saved

$110,000+ in annual productivity gains

Plus: risks caught 3 weeks early · institutional knowledge retained · eliminated team dependencies · measurably improving accuracy

How the Free Pilot Works¶

Zero risk. We automate one workflow in 2-3 weeks — you measure the results before spending anything.

1

Pick One Workflow

Which reporting workflow wastes the most time? Weekly status? Ad-hoc data requests? Executive reports? We start with the one that hurts most.

2

We Automate It

In 2-3 weeks, we deploy the agent on your machine, connect it to your tools, configure your Shared Brain, and automate the workflow end-to-end.

3

You Measure Results

No cost. No commitment. You measure real time saved, accuracy scores, and evolution learning rate. If it works — we expand. If not — zero money wasted.

Your Data Stays on Your Machine¶

🔒 Nothing leaves. Ever.

❌ Never

Project Data (Commits, PRs, Tickets)

❌ Never

Database Queries & Results

❌ Never

Chat Conversations

❌ Never

Agent Memory & Knowledge

❌ Never

API Keys & Credentials

Using **Ollama**? Even your AI conversations stay 100% on your machine. Zero cloud, zero tracking. [Learn more about our security model →](security.md)

Built For Your Role¶

🏗️

CTOs & VP Engineering

Portfolio risk visibility → proactive decisions, not fire drills

👔

Engineering Managers

Automate status gathering → real-time project health

📊

Data & Analytics Leaders

Eliminate ad-hoc request backlog → self-service in 3 sec

📋

Technical PMs

3 report variants → one click, CTO/CFO/PMO in seconds

Free AI Automation Pilot¶

No Cost. No Commitment. Just Results.

We pick ONE workflow — the one that wastes the most time — and automate it in 2-3 weeks. You see a real, working result before spending anything. Our only ask: honest feedback.

✅ Full deployment on your machine

✅ Connected to your tools — GitHub, Jira, Slack, or databases

✅ Shared Brain configured with your team's KPIs & glossary

✅ Measurable results — time saved, accuracy score, evolution rate

✅ No cost, no commitment — zero risk

Request Your Free Pilot ✉️ Download & Try Yourself ⬇️

📧 info@unicolab.ai · 💬 DM us on LinkedIn · 🌐 unicolab.ai

For Technical Teams: Under the Hood

The AgentOS Platform¶

AgentOS is a composable, local-first AI agent engine built in Go 1.23 with a React 19 frontend. Each copilot is a specialized "pack" — a domain configuration that includes tools, dashboards, AI persona, and workflows.

Architecture:

┌─────────────────────────────────────────────────────┐
│                   AgentOS Engine                     │
│                                                      │
│  ┌──────────┐  ┌──────────────┐  ┌──────────────┐  │
│  │ Sidebar  │  │   Dashboard  │  │   Copilot    │  │
│  │          │  │   (Bento)    │  │   (Chat)     │  │
│  │ Modules  │  │   24 cards   │  │  Streaming   │  │
│  │ per pack │  │  per pack    │  │  + Tools     │  │
│  └──────────┘  └──────────────┘  └──────────────┘  │
│                                                      │
│  ┌────────────────────────────────────────────────┐  │
│  │         Intelligence Layer (Go)                │  │
│  │  Adaptive Memory · Knowledge Graph · Steering │  │
│  │  Shared Brain · Error Patterns · Tool Wisdom  │  │
│  │  Performance Tracker · User Adaptation         │  │
│  └────────────────────────────────────────────────┘  │
│                                                      │
│  ┌────────────────────────────────────────────────┐  │
│  │          Tool Gateway                          │  │
│  │  GitHub · Jira · Slack · Database · Shell     │  │
│  │  Gmail · Calendar · Docs · HTTP · Web Search  │  │
│  └────────────────────────────────────────────────┘  │
└─────────────────────────────────────────────────────┘

Available Packs:

Pack	Persona	Domain	Status
`aiflow-pm`	🎩 Jean-Pierre	Project Management	✅ Production
`michelle`	🔬 Michelle	Analytics Intelligence	✅ Production
`edith`	🥐 Édith	Sales Intelligence / CRM	✅ Available
`freelancer`	💼 Yvette	Freelancer PM	✅ Available
`office`	🏢 Office	Productivity	✅ Available
`retail-ops`	🏪 Retail	Operations	✅ Available

AI Providers:

Provider	Type	Cost
Ollama	100% local	Free
OpenAI	Cloud API	Per-token
Anthropic	Cloud API	Per-token
Google Gemini	Cloud API	Per-token

Tech Stack: Go 1.23 · React 19 · TypeScript · Vite 7 · SQLite · SSE Streaming · Glassmorphism UI

Quick Install:

curl -fsSL https://unicolab.github.io/agentos/install.sh | sh
agentos serve
# Open http://localhost:18080

Full Documentation → API Reference →

Meet Jean-Pierre & Michelle.AI That Actually Works.¶

What They Automate For You¶

🎩 Jean-Pierre

🔬 Michelle

See the Results¶

Project Reporting Automation

Analytics Intelligence

Why They Don't Hallucinate¶

They Evolve

They Validate

They Stay Private

9 Layers of Anti-Hallucination¶

"But I Already Have Claude…"¶

The Time You're Losing Today¶

PMs: 45 min of chaos every morning¶

Data teams: buried in lookups¶

Knowledge walks out the door¶

AI chatbots hallucinate numbers¶

The Time You Get Back¶

How the Free Pilot Works¶

Pick One Workflow

We Automate It

You Measure Results

Your Data Stays on Your Machine¶

Built For Your Role¶

CTOs & VP Engineering

Engineering Managers

Data & Analytics Leaders

Technical PMs

Free AI Automation Pilot¶

The AgentOS Platform¶

Meet Jean-Pierre & Michelle.
AI That Actually Works.¶