Builder Handbook
For builders advancing from Forge Sprint 02 · ProcessHQ Hackathon Edition · Sprint Day Saturday 4 July 2026 · NMG Gurugram · 09:00 to 20:00 IST
01 Welcome: From Forge 2 to Forge 3
You made it. In Forge 2, you built a multi-agent AI system using Hermes Agent and OpenClaw. You learned orchestration, Slack-based coordination, model routing and CI/CD for agent-generated code.
Forge 3 is where those skills meet a real product.
ProcessHQ.AI is not a hackathon demo. It is a live SaaS platform with 56+ enterprise demos, real paying customers and a product roadmap full of features that businesses are demanding right now. The feature you build at Forge 3 has a real chance of shipping to production.
What Changed from Forge 2
| Forge 2 | Forge 3 |
|---|---|
| Any problem domain | ProcessHQ-specific features |
| Demo output sufficient | Production-quality code expected |
| 100-point rubric | 150-point rubric (harder, more rewarding) |
| Small prizes | Significantly larger, revealing soon |
| Apprenticeship/internship | Full-time engineering offers for Champion |
| Your code stays in your repo | Champion's feature ships to production |
What You Already Have from Forge 2
Your multi-agent system from Forge 2 is your weapon. Bring it ready:
- ☐ Hermes Agent configured and working (orchestrator)
- ☐ OpenClaw configured and working (coding agent)
- ☐ Slack workspace with channels set up
- ☐ Model routing configured (planning model + execution model)
- ☐ CI/CD pipeline working (GitHub Actions or equivalent)
- ☐ agent-log.md capture working
If any of the above is broken, fix it BEFORE Sprint Day. Forge 3 is not the place to debug your agent setup. You should spend 100% of your 6 hours building the ProcessHQ feature.
02 What is ProcessHQ.AI, Product Deep Dive
The One-Liner
ProcessHQ.AI is a Process Operating System that replaces Google Sheets for business operations, task management, flow management, inventory management and bill of materials in one platform.
The Origin Story
ProcessHQ was built by Rohit Maheshwari (NMG Partner, MSc Business Systems from City University London, ex-PlayStation/20th Century Fox) after Google Forms and Excel broke at scale while trying to manage NMG's own 150-person operation. The product grew out of Toyota Process Management (TPM) principles applied to Indian mid-market businesses.
Who Uses It
ICP (Ideal Customer Profile): Organizations with 30-250 users, specifically NCR mid-market manufacturers or multi-branch retailers.
Industries (from 56 demos):
- › Manufacturing (18+ demos): Steel, plastics, alloys, garments, food processing, chemicals, furniture, stainless steel, power tools
- › Trading & Distribution (8+ demos): Steel distribution (Tata Steel distributor), FMCG, exports
- › Education (3 demos): K-12 schools (Khaitan Schools with 150 FMSs), vocational training
- › Real Estate & Construction (4 demos): Multi-site operations, vendor/client portals
- › Services & Advisory (5+ demos): CA firms, business coaching, financial services
The Technology
- › Backend: Laravel (PHP)
- › Database: MySQL
- › Hosting: AWS
- › Frontend: Web-based dashboard
- › Mobile: Under development (iOS + Android)
The Business Model
- › SaaS subscription pricing (per-user model)
- › Serves organizations ranging from 10 to 250+ users
- › Volume pricing available for larger teams
Visit and study the product at processhq.ai.
03 ProcessHQ Modules Explained
PMS (Project Management System)
The task management layer. Every task has an owner (doer), a delegator, a deadline and a checklist.
Key features:
- › Task creation with delegation (up to 3 levels: executive to manager to director)
- › Recurring tasks with frequency control (daily, weekly, monthly)
- › Help ticket system with escalation rules
- › Checklists with completion tracking
- › Voice notes for task instructions
- › Performance dashboards (standard view, grid view, 10,000-feet view)
- › Reports/MIS with department-wise breakdowns
What customers ask for that doesn't exist yet:
- + Graphical trend dashboards (week-over-week with % arrows)
- + Score-based employee filtering (show below 50%)
- + Holiday calendar auto-rescheduling
- + AI resolution from historical ticket data
FMS (Flow Management System)
The process automation layer. A Flow is a multi-step process with defined stages, each assigned to a doer.
Flow Types:
- › Direct FMS: Linear, step-by-step (A to B to C to D)
- › Loop FMS: Includes rejection/rework loops (if step C fails QC, loop back to step B)
- › Staggered FMS: Partial deliveries across stages
- › Conditional FMS: Diamond-box decision points (pass/fail gates)
Key features:
- › Auto-generated flowchart visualization
- › Bottleneck graphs showing delay severity (0% to 75%+)
- › Step-level notifications (email, planned WhatsApp)
- › Document proof mandate per step
- › Favorites system for frequently used flows
What customers ask for that doesn't exist yet:
- + Visual drag-and-drop flow builder (draw process FIRST, then system generates steps)
- + Cross-FMS data linking (data from one flow automatically feeding into another)
- + External user access (clients/vendors view their own flow status)
- + Process GPT (describe process in plain English, AI generates the FMS)
IMS (Inventory Management System)
Stock tracking for manufacturing and trading businesses.
Key features:
- › Product master with categories
- › Stock-in/stock-out tracking
- › Warehouse management (location/bin tracking)
- › Reorder alerts
What customers ask for that doesn't exist yet:
- + Vendor-product linkage (which vendor supplies which product)
- + Committed vs live stock (what's ordered but not received)
- + Min/max reorder levels with auto-PO triggers
- + Purchase price history per vendor
- + Integration with purchase FMS
Critical feedback from Sanjit Mutha: "Study Odoo's Purchase + Inventory modules as a reference."
BOM (Bill of Materials)
Component breakdown for manufacturing.
Key features:
- › Component-level breakdown for products
- › ~80% complete module
What customers ask for that doesn't exist yet:
- + Combi Master for shared/common components across products
- + Shortage calculation and reorder sheet
- + Auto-PO generation from BOM shortages
- + BOM to IMS to Purchase FMS linkage (the full manufacturing chain)
04 Customer Pain Points
From 56 real demo transcripts. These are REAL quotes and patterns from actual ProcessHQ demo conversations. Understanding these deeply is part of your judging score.
The Top 5 Pain Points (Universal)
1. Google Sheets Doesn't Scale (Every single demo)
- › Sheets hang and load slowly with large data
- › Cross-linked sheets break and corrupt
- › No graphical dashboards, just numbers
- › Manual MIS creation with copy-paste from multiple sources
- › Examples: Deepak Jain (Delta Exports), Atul Garg (Oddy Paper), Vedant Khaitan (Khaitan Schools)
2. No Accountability (35+ demos)
- › Tasks scattered across WhatsApp, email, verbal meetings, Google Sheets
- › Verbal commitments get lost. No audit trail.
- › "55% of time wasted on follow-ups" (Deborah's demo)
- › "35% revenue loss from communication breakdowns" (cited statistic)
3. DME/Techie Dependency (15+ demos)
- › Companies pay 30-50K/month for one person to maintain their Google Sheets ecosystem
- › When that person leaves, the system breaks
- › Examples: Vedant (DME team for Looker Studio), Punjab BG (DME dependency), Agronic (one techie built everything)
4. No ERP Integration (~22 demos, 40% stall)
- › Companies already have Tally, Busy, Oracle, SAP, MS Business Central for accounting
- › ProcessHQ requires manual dual entry, prospects refuse
- › This is the single biggest dealbreaker
5. Manufacturing Needs Not Met (18+ demos)
- › BOM not fully integrated with IMS
- › No production scheduling
- › No warehouse bin/rack/location structure
- › Factory workers need mobile access but don't have email
Industry-Specific Pain Points
| Industry | Specific Pain | Example Prospect |
|---|---|---|
| Manufacturing | BOM + IMS + Purchase FMS not linked. Production-office coordination gaps. | Padam (large-scale orders), Armaan (furniture), Evergreen (3 branches) |
| Education | 15-min refresh lag on Looker Studio. Cannot scale to 50 schools. | Vedant Khaitan (150 FMSs, thousands of checkpoints) |
| Trading | 150 vendors, 20 clients, 5-7 sites each. No structured process visibility. | Shashwat (electrical distribution), Daga (Tata Steel, 160 users) |
| Construction | Need cross-FMS linking (site attendance to DPR cost calculation). Vendor/client portals. | Punjab BG (tried Jira, Trello, ClickUp, Monday, none fit) |
| Food Processing | 30 FMSs running well on Google Sheets + Python. Floor workers update via proxy. | Agronic (350 staff, 6 years on Sheets), Saket (11 plants, 50+ team) |
| Garments | 20-25 day lead times. Everything on paper then manually digitized. ERP too slow. | Meenu Creation (Zara supplier, 15-20 floor supervisors) |
05 The 7 Challenge Tracks
You will be assigned ONE track on Sprint morning. Study ALL tracks so you can hit the ground running regardless of assignment.
Process GPT
- › Build: Natural-language interface to FMS generation
- › Customer voice: "I should be able to describe my process in English and the system should create it"
- › Tech: NLP intent parser + flow generator + validation engine
- › Agents needed: 3 (parser, generator, validator)
- › Difficulty: Hard (requires understanding FMS structure deeply)
Capacity IQ
- › Build: AI capacity planning from task history data
- › Customer voice: "Show me who is overloaded before the SLA breaks" (Vidur, Rahul Agarwal)
- › Tech: Data pipeline + ML prediction model + dashboard
- › Agents needed: 3 (data, model, dashboard)
- › Difficulty: Hard (requires ML/analytics skills)
MIS/Analytics Engine
- › Build: Graphical dashboards with trend graphs, % changes, department views
- › Customer voice: "I want to see week-over-week trends with arrows, not just numbers" (Vidur Nath)
- › Tech: Aggregation pipeline + Chart.js/D3 renderer + PDF/Excel export
- › Agents needed: 3 (aggregation, rendering, export)
- › Difficulty: Medium (well-defined output, clear customer expectations)
ERP Bridge Connector
- › Build: Bi-directional connector to ONE ERP (Tally/Zoho/BUSY)
- › Customer voice: "Will I still need to copy-paste?" (Gurjeet, frustrated)
- › Tech: ProcessHQ-side API + ERP-side API + data mapping + conflict resolution
- › Agents needed: 3 to 4 (ProcessHQ connector, ERP connector, mapper, conflict handler)
- › Difficulty: Hard (ERP APIs are complex and poorly documented)
Visual Flow Builder
- › Build: Drag-and-drop flowchart to auto-generated FMS
- › Customer voice: "I want to draw the process first, then the system creates the steps" (Gaurav)
- › Tech: React DnD canvas + flow-to-backend compiler + validation logic
- › Agents needed: 3 (canvas, compiler, validator)
- › Difficulty: Hard (frontend + backend + domain logic)
Help Ticket System v2
- › Build: Auto-routing + SLA tracking + escalation + AI suggestions
- › Customer voice: "Can AI suggest solutions from past tickets?" (Gaurav)
- › Tech: Rule engine + notification system + AI suggestion engine + dashboard
- › Agents needed: 3 to 4 (router, escalation, AI engine, dashboard)
- › Difficulty: Medium (well-scoped, clear requirements)
The Executive Co-Pilot
- › Build: Ingest messy comms to structured accountability brief
- › Customer voice: "Work is scattered across WhatsApp, email, calls, meetings" (nearly every demo)
- › Tech: Multi-format ingester + entity extractor + report generator
- › Agents needed: 3 (ingester, extractor, reporter)
- › Difficulty: Medium (text processing + structured output)
06 Pre-Sprint: Come Prepared
Your Multi-Agent System (from Forge 2)
- ☐ Hermes Agent fully working and tested
- ☐ OpenClaw fully working and tested
- ☐ Slack workspace with channels configured
- ☐ Model routing working (planning + execution models)
- ☐ CI/CD pipeline functional
- ☐ Can spin up a 3-agent system and have them coordinate within 15 minutes
ProcessHQ Domain Knowledge
- ☐ Read this entire handbook (you are reading it now)
- ☐ Understand all 4 modules (PMS, FMS, IMS, BOM)
- ☐ Understand the 4 FMS flow types (Direct, Loop, Staggered, Conditional)
- ☐ Read the customer pain points section (Section 04), this is tested in judging
- ☐ Study all 7 challenge tracks, you will be assigned ONE randomly
- ☐ Visit processhq.ai and study the product
Hardware & Software
- ☐ Laptop with 8+ GB RAM, charger, government ID, headphones
- ☐ All Forge 2 tools installed and verified
- ☐ GitHub account ready
- ☐ Loom account (free) for walkthrough video
07 Learning Curriculum, ProcessHQ Domain
In addition to your Forge 2 skills (which should already be solid), study these ProcessHQ-specific materials:
Module 1: Understand ProcessHQ Product (Required)
| Resource | Type | Time |
|---|---|---|
| ProcessHQ.AI website | Website | 30m |
| ProcessHQ demo videos (provided to shortlisted builders) | Video | 2h |
| This handbook, Sections 02, 03, 04 (modules + pain points) | Document | 1h |
Module 2: Understand the Customer (Required)
| Resource | Type | Time |
|---|---|---|
| Customer pain points summary (Section 04 of this handbook) | Document | 30m |
| Selected demo transcript excerpts (provided to shortlisted builders) | Transcripts | 1h |
| Industry breakdown and prospect evidence | Document | 30m |
Module 3: Study the 7 Tracks (Required)
| Resource | Type | Time |
|---|---|---|
| Section 05 of this handbook, all 7 tracks in detail | Document | 1h |
| Related open-source implementations (research on your own) | Research | 2h |
Module 4: Refresh Multi-Agent Skills (If Needed)
| Resource | Type | Time |
|---|---|---|
| Your Forge 2 agent system, verify everything still works | Hands-on | 1h |
| Forge 2 Builder Handbook, Modules 4-7 if rusty | Document | 1h |
Total Pre-Sprint Study: ~10 hours
(On top of the ~14 hours you already invested for Forge 2)
08 Sprint Day Challenge Structure
What You Receive at 09:50 AM
- 1 Your track assignment (one of Track A through G)
- 2 ProcessHQ starter repo or sandbox environment
- 3 Real customer data (anonymized) relevant to your track
- 4 Customer quotes from actual demo transcripts, the exact words prospects used to describe their need
- 5 ProcessHQ API documentation for the modules relevant to your track
Architecture Requirements (Mandatory)
- ☐ Hermes Agent as orchestrator
- ☐ OpenClaw as at least one coding agent
- ☐ Slack channels showing all coordination
- ☐ Automated tests running on the ProcessHQ codebase
- ☐ Human approval step before final output
- ☐ Minimum 3 agents with distinct roles
Required Outputs
- 1 Working feature that runs against ProcessHQ test suite
- 2 Public GitHub repo with incremental commits + agent-log.md
- 3 Slack export showing agent coordination
- 4 README with architecture diagram
- 5 Loom video (max 7 minutes)
- 6 1-page customer pitch: "How I would demo this feature to a prospect"
09 Sprint Day Schedule
| Time | Block | Details |
|---|---|---|
| 09:00-09:30 | Check-in | ID check, verify agent systems work, ProcessHQ repo access |
| 09:30-09:50 | ProcessHQ Deep Dive | Rohit Maheshwari walks through the product live. Q&A with ProcessHQ engineering. |
| 09:50 | Track assignment | Each builder gets their specific track + customer data |
| 10:00 | Build starts | 6-hour window. ProcessHQ engineers available as mentors. |
| 10:30 | Connection check | Verify agents connected and communicating |
| 13:00-14:00 | Lunch | Eat with ProcessHQ team and judges |
| 14:00 | Mid-sprint review | ProcessHQ PM reviews progress per track |
| 15:30 | Scope or cut | Core feature must work. Drop polish. |
| 16:30 | 30-min warning | Test against ProcessHQ suite. Record Loom. Write customer pitch. |
| 17:00 | Submission lock | GitHub frozen. Second customer dataset released. |
| 17:00-17:30 | Auto-tester | Runs against ProcessHQ test suite + surprise data |
| 17:30-19:30 | Demos | 7 min: 4-min demo + 3-min Q&A from ProcessHQ team + judges |
| 19:30 | Deliberation | ProcessHQ engineering evaluates production-readiness |
| 20:00 | Champions announced | Major prizes extended on the night |
10 Judging Rubric (150 Points)
| Criterion | Points | Description |
|---|---|---|
| Feature Completeness | 30 | Does it work end-to-end? Does it solve the customer problem? |
| Production Readiness | 25 | Could this ship? Code quality, error handling, edge cases, tests. ProcessHQ engineering team scores this. |
| Multi-Agent Architecture | 25 | Clean Hermes/OpenClaw/Slack setup. Model routing. Improvement from your Forge 2 system. |
| Customer Understanding | 20 | Does your 1-page pitch show you understand the customer? Does the feature address specific pain points from demos? |
| Domain Knowledge | 15 | Correct use of PMS/FMS/IMS/BOM concepts. Understanding of flow types, task delegation, bottleneck analysis. |
| Innovation | 15 | Went beyond the brief? Self-improving agents that learn ProcessHQ patterns? Unexpected valuable additions? |
| Presentation | 10 | Demo quality, README, architecture diagram, Loom video. |
| Agent Logs | 10 | Complete agent-log.md and Slack evidence. |
| TOTAL | 150 |
11 Prizes & Career Outcomes
What Every Participant Gets
| Outcome | Details |
|---|---|
| Cash Prizes | Revealing soon, to be announced before Sprint Day |
| NMG Offer | Top performers receive NMG apprenticeship, internship or full-time offers |
| Certificate of Participation | Official nmg.labs Forge Sprint 03 certificate for all participants |
| NMG Labs Skool Membership | Lifetime membership to the NMG Labs Skool community |
Place-Specific Outcomes
| Place | Prize | Additional |
|---|---|---|
| Champion (1st) | Cash (revealing soon) + NMG offer + Certificate + Skool | Feature fast-tracked to ProcessHQ production. Name in changelog. |
| Runner-up (2nd) | Cash (revealing soon) + NMG offer + Certificate + Skool | Feature evaluated for production. |
| 3rd Place | Cash (revealing soon) + NMG offer + Certificate + Skool | Feature reviewed by PM. |
| Track Bests (A-G) | Track-specific recognition + Certificate + Skool | Best per track awarded separately |
| Top 10 | Certificate + Skool + NMG interview fast-track | Priority for future Forges |
| All Participants | Certificate + NMG Labs Skool membership | Standing invitation to future sprints |
12 Rules & Disqualification
Tie-Breaker Rules
- 1 Production readiness score decides
- 2 If still tied: ProcessHQ engineering team picks their preferred implementation
- 3 If still tied: feature that handles the surprise dataset better wins
Intellectual Property
Your code, your IP. NMG retains a non-exclusive license to publish a case study about your build (with credit to you). The Champion's feature may be fast-tracked to ProcessHQ production, in which case you work directly with the engineering team on the integration.
Code of Conduct
Zero tolerance for plagiarism, harassment or AI-generated submissions misrepresented as your own work. We expect respect, intellectual honesty and humour under pressure. Violations = disqualification.
8 Standard Triggers (Same as Forge 2)
- 1 Private repo at deadline
- 2 Missing agent-log.md or Slack evidence
- 3 No Hermes Agent as orchestrator
- 4 Single giant commit
- 5 Auto-tester failure within 5 minutes
- 6 Cannot explain any code block
- 7 75%+ code duplication
- 8 Conduct violation
2 Additional Forge 3 Triggers
- 9 Feature does not connect to ProcessHQ architecture at all
- 10 No evidence of understanding customer pain points (empty customer pitch or generic content)
Forge Captain: Ayush Gupta · labs@nmgdigital.com
ProcessHQ questions: rohit@nmgtechnologies.com
Forge Sprint 03 · ProcessHQ Hackathon · 4 July 2026 · NMG Gurugram. An initiative of nmg.labs by NMG Technologies.