Build real features for a live SaaS. Ship in 6 hours.
You learned the weapon in Forge 2. Now use it to build something real.
In Forge 2 you built a multi-agent AI system with Hermes Agent, OpenClaw and Slack. Forge 3 is where you deploy those skills on ProcessHQ.AI, a live SaaS platform built by NMG with a real codebase, a real user base and a roadmap of features customers are asking for right now.
· Editions
Forge 3 runs as editions of the same challenge. Pick the edition you want. Open editions link straight to registration, completed editions link to their results.
· Leaderboard
Scores only, the way Forge runs. Switch between the all-time best across every Forge 3 edition and a single edition ranking.
Why Forge 3 prizes are bigger
Each Forge raises the stakes. Forge 1 built a skill for a solo agent. Forge 2 built a multi-agent system as a demo. Forge 3 builds a real product feature for a live SaaS, code that has a genuine chance of shipping to paying customers.
| Comparison | Forge 1, Claude Code | Forge 2, Hermes + OpenClaw | Forge 3, ProcessHQ |
|---|---|---|---|
| What you build | A skill for a solo agent | A multi-agent system (demo) | A real product feature for a live SaaS |
| Impact | Personal portfolio | Architecture proof of concept | Could ship to 56+ enterprise customers |
| Difficulty | Solo agent, 6 hours | Multi-agent coordination | Multi-agent plus domain knowledge plus real codebase |
| Prize pool | Apprenticeship plus cash | Bigger cash plus Forge 3 access | Significantly larger, revealing soon |
Not demos. Actual features from a live roadmap.
The Champion's feature is fast-tracked to the ProcessHQ roadmap.
Forge 2 established your multi-agent foundation.
What you build maps directly to revenue and customer retention.
What is ProcessHQ.AI?
ProcessHQ.AI is a Process Operating System that replaces Google Sheets for business operations. It runs on four core modules and is already live with real customers across 15+ industries.
The four core modules
| Module | What it does | Key features |
|---|---|---|
| PMS Project Management | Task management with delegation, checklists and accountability | Recurring tasks, help tickets, escalation, dashboards, voice notes, WhatsApp notifications |
| FMS Flow Management | Process automation with step by step workflows | Loop, Direct and Staggered flows, bottleneck graphs, auto-generated flowcharts, conditional logic |
| IMS Inventory Management | Stock tracking, reorder alerts and warehouse management | Vendor master, purchase FMS, GRN tracking, stock alerts |
| BOM Bill of Materials | Component breakdown for manufacturing | Combi Master, shortage calculation, auto-PO generation |
The market
Industries: Manufacturing, Trading, Food Processing, Real Estate, Education, Services, Automotive, Garments, Chemicals, Jewelry, IT/SaaS and Telecom.
ICP: 30 to 250 user organizations, NCR mid-market manufacturers and multi-branch retailers. Every single prospect currently runs on Google Sheets, which makes the migration path universal.
The top five problems it solves
| # | Pain point | What goes wrong today |
|---|---|---|
| 1 | Google Sheets does not scale | Slow loading, data fragmentation, manual MIS, sheets break when cross-linked |
| 2 | No accountability | Tasks scattered across WhatsApp, email, verbal meetings and Google Sheets |
| 3 | DME or techie dependency | Companies pay 30K to 50K per month for one person to maintain their Google Sheets ecosystem |
| 4 | No visibility | Managers cannot see bottlenecks, overdue tasks or team performance without manual reports |
| 5 | Manual MIS creation | Every report is copy-pasted from multiple sources with no trend graphs or automated analytics |
Use case tracks, one assigned per builder
The exact track assignment happens on Sprint morning. You will receive ONE of the seven tracks below, drawn straight from the ProcessHQ roadmap and 56 demo transcripts. Each comes with real customer evidence and a suggested multi-agent approach.
Process GPT
Build: a natural-language interface where a user describes a business process in plain English and the system generates a complete FMS with steps, checklists, roles and conditional logic.
Customer evidence: on the ProcessHQ roadmap since May 2026. Chat-driven FMS creation is a stated product goal.
Capacity IQ
Build: an AI module that watches work in progress across teams and flags overloaded teammates before SLAs break, using task history, resource utilization and bottleneck data to predict capacity.
Customer evidence: on the roadmap. Vidur Nath and Rahul Agarwal asked for score-based employee filtering and benchmark gap analysis.
MIS / Analytics Engine with Trend Graphs
Build: the graphical analytics layer with week over week trends and percentage changes, department-wise views, score-based employee filtering, benchmark gap analysis and PDF or Excel export.
Customer evidence: 22+ demos requested this. Vidur wanted trend graphs with up and down arrows. The current system is numbers only.
ERP Bridge Connector
Build: a bi-directional connector between ProcessHQ and ONE ERP system (Tally OR Zoho OR BUSY). It must sync tasks, inventory and purchase orders in both directions.
Customer evidence: around 22 demos plus a 40% dealbreaker. Raghav has Oracle with a dual entry problem. Gurjeet has Busy. Daga has MS Business Central.
Visual Flow Builder
Build: a drag and drop flowchart designer where users draw the process first, then the system generates the underlying FMS steps. Include conditional diamond-box decision points for QC pass or fail gates.
Customer evidence: 10+ demos. Gaurav asked for "flow chart first" design. Dilip needed diamond-box approval gates.
Help Ticket System v2
Build: an intelligent help ticket system with auto-routing by category, SLA tracking, multi-tier escalation from executive to manager to director and AI resolution suggestions from historical ticket data.
Customer evidence: 21 demos. Gaurav wanted AI auto-resolve from history. Mukul had 112 overdue from 9 templates.
The Executive Co-Pilot
Build: a system that ingests a week of messy communications (chat logs, voice transcripts, meeting notes and email threads) and produces a structured Friday accountability brief, the MD's weekly operating system.
Customer evidence: multiple prospects described work scattered across WhatsApp, email and verbal meetings. A 55% time waste stat was cited in Deborah's demo.
How Forge 3 works
Eligibility
| Who | Status |
|---|---|
| Forge 2 Champions, 1st, 2nd and 3rd | Guaranteed seat. Dedicated NMG mentor assigned. |
| Forge 2 Top 10 | Guaranteed seat. Priority support. |
| Forge 2 participants | Can apply. Selection based on Forge 2 performance. |
| Fresh applicants | Can apply, but Forge 2 alumni get priority. |
On Sprint morning you receive one of the ProcessHQ use cases assigned as your challenge, the ProcessHQ codebase (or a sandbox starter repo matching the architecture), real anonymized customer data from actual demos and customer quotes from transcripts showing exactly what prospects asked for. You then build a working feature using your Hermes plus OpenClaw multi-agent system from Forge 2.
What you bring from Forge 2
| Forge 2 skill | How it applies in Forge 3 |
|---|---|
| Hermes orchestrator | Decomposes the ProcessHQ feature into sprint tasks and assigns to agents |
| OpenClaw coding agents | Write the actual ProcessHQ feature code |
| Slack coordination | Monitor agent progress and intervene when agents go off track |
| Model routing | Opus or Kimi for architecture decisions, DeepSeek for bulk coding |
| CI/CD pipeline | Automated tests on the ProcessHQ codebase with quality gates |
| Self-improving skills | Agents learn from ProcessHQ domain errors and build reusable skills |
Sprint day schedule, 4 July 2026
| Time (IST) | Block | Details |
|---|---|---|
| 09:00 to 09:30 | Check-in | Photo ID, stack verification, ProcessHQ codebase access granted |
| 09:30 to 09:50 | ProcessHQ Deep Dive | Live walkthrough of ProcessHQ by Rohit Maheshwari (architect). Product overview, codebase orientation, API documentation. |
| 09:50 | Track assignment | Each builder receives their specific use case track plus customer data |
| 10:00 | Build window opens | 6-hour countdown. ProcessHQ engineering team available as mentors. |
| 10:30 | Architecture checkpoint | Verify multi-agent system connected. Confirm understanding of the track. |
| 13:00 to 14:00 | Catered lunch | Builders eat with the ProcessHQ engineering team and judges |
| 14:00 | Mid-sprint review | Forge Captain plus ProcessHQ PM review progress |
| 15:30 | Scope or cut | Stabilize the core feature. Drop polish items that will not ship. |
| 16:30 | 30-minute warning | Final testing against the ProcessHQ test suite. Record Loom. |
| 17:00 | Submission lock | GitHub frozen. Second customer dataset released for auto-testing. |
| 17:00 to 17:30 | Auto-tester | Judges run builds against the ProcessHQ test suite plus the second dataset |
| 17:30 to 19:30 | Finalist demos | 7 minutes each: 4-min demo plus 3-min Q&A from ProcessHQ team and judges |
| 19:30 | Judging deliberation | ProcessHQ engineering team evaluates production-readiness |
| 20:00 | Champions announced | Major prizes plus NMG engineering offers extended |
Judging criteria, 150 points total
The rubric is larger than Forge 2 because the bar is higher. The ProcessHQ engineering team evaluates production-readiness directly.
| Criterion | Points | What judges look for |
|---|---|---|
| Feature Completeness | 30 | Does the feature work end to end? Does it solve the customer problem described in the track? |
| Production Readiness | 25 | Could this ship to ProcessHQ's codebase? Code quality, error handling, edge cases and tests. The ProcessHQ engineering team evaluates. |
| Multi-Agent Architecture | 25 | Clean orchestrator and worker separation. Slack communication visible. Model routing demonstrated. Improvement from your Forge 2 setup. |
| Customer Understanding | 20 | Did you deeply understand the customer need? Does your solution address the specific pain points from demo transcripts? |
| Domain Knowledge | 15 | Understanding of ProcessHQ's modules (PMS, FMS, IMS, BOM). Correct use of domain concepts. |
| Innovation | 15 | Did you go beyond the brief? Add something unexpected that makes the feature better? Self-improving agents that learn ProcessHQ domain patterns? |
| Presentation | 10 | Demo quality, README, architecture diagram and Loom video. Can you sell the feature to a prospect? |
| Agent Logs & Slack Evidence | 10 | Complete agent-log.md and Slack export showing the full coordination story. |
| TOTAL | 150 |
Tie-breaker
- Production readiness score decides.
- If still tied, the ProcessHQ engineering team picks their preferred implementation.
- If still tied, the feature that handles the surprise dataset better wins.
Prizes and 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: a private builder network, resources, mentorship and priority access to future Forges |
Place-specific outcomes
| Place | Prize | Additional |
|---|---|---|
| Champion, 1st | Cash prize (revealing soon) plus NMG offer plus Certificate plus NMG Labs Skool | Feature fast-tracked for ProcessHQ production. Your name in the product changelog. Dedicated mentor from ProcessHQ engineering. |
| Runner-up, 2nd | Cash prize (revealing soon) plus NMG offer plus Certificate plus NMG Labs Skool | Feature evaluated for production inclusion. |
| 3rd Place | Cash prize (revealing soon) plus NMG offer plus Certificate plus NMG Labs Skool | Feature reviewed by the ProcessHQ PM. Priority for the next product hackathon. |
| Track Bests, A to G | Track-specific recognition plus Certificate plus NMG Labs Skool | Best implementation per track awarded separately. |
| Top 10 Finalists | Certificate plus NMG Labs Skool plus NMG interview fast-track | Priority for all future Forge sprints. |
| All Participants | Certificate of Participation plus NMG Labs Skool membership | Standing invitation to future sprints. |
Your feature in production
The Champion's feature gets fast-tracked into the ProcessHQ production roadmap. That means your code ships to real customers, your name goes in the product changelog, you work directly with the ProcessHQ engineering team to production-ready it and you have a live deployed feature that 56+ enterprises will use.
This is not a certificate. This is not a project for your resume. This is real software shipping to real businesses.
The stack
Required tools
| Tool | Role | Why |
|---|---|---|
| Hermes Agent | Orchestrator | Self-improving agent that plans, decomposes tasks and coordinates workers |
| OpenClaw | Coding Agent | Multi-channel gateway for LLM coding agents. Writes the actual ProcessHQ code. |
| Slack (free workspace) | Communication | All agent communication. Human in the loop. Matches production patterns. |
| GitHub | Version Control | Public repo with incremental commits. Required for judging. |
| ProcessHQ Starter Repo | Codebase | Provided on Sprint morning. Sandbox or starter matching the ProcessHQ architecture. |
Model options, all free
| Model | Provider | Best for | Free tier |
|---|---|---|---|
| Kimi K2 Instruct | Groq | Orchestration plus planning | ~1,000 req/day |
| DeepSeek V3 | DeepSeek API | Bulk coding | Free tier available |
| Qwen2.5-Coder 7B | Ollama (local) | Offline coding | Unlimited (local) |
Rules
Eligibility
- Forge 2 Champions, 1st to 3rd: automatic entry.
- Forge 2 Top 10: guaranteed entry.
- Forge 2 participants: application based, with priority over fresh applicants.
- Fresh applicants: must demonstrate equivalent multi-agent skills via portfolio review.
Build requirements
- Must use Hermes Agent as orchestrator.
- Must use OpenClaw as at least one coding agent.
- Must use Slack for agent communication.
- Must work within the ProcessHQ architecture (starter repo provided).
- Public GitHub repo with incremental commits required.
Submission requirements
- Public GitHub repo with agent-log.md.
- Slack export showing agent communication.
- README with architecture diagram.
- Loom video walkthrough, max 7 minutes.
- Working feature that runs against the ProcessHQ test suite.
- 1-page customer pitch: how you would demo this feature to a prospect.
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 equal disqualification. Read the full Terms & Conditions and Code of Conduct.
Disqualification triggers
The same 8 standard Forge 2 triggers apply, plus 2 extra triggers specific to Forge 3.
| # | Trigger |
|---|---|
| 1 | No public GitHub repo or no incremental commit history |
| 2 | Missing agent-log.md or no Slack export of agent communication |
| 3 | No working multi-agent system (single agent only) |
| 4 | Hermes Agent not used as orchestrator |
| 5 | OpenClaw not used as a coding agent |
| 6 | Plagiarism or misrepresented AI-generated work |
| 7 | Harassment or any code of conduct violation |
| 8 | Late submission after the GitHub freeze |
| 9 | Feature does not connect to the ProcessHQ architecture at all |
| 10 | No evidence of understanding customer pain points from demo transcripts |
Build something that ships. Register for Forge 3 · Edition 1.
Bring your multi-agent team and pick up a real feature for a live SaaS. Forge 3 is an individual competition. Register below and we will email you the check-in details before Sprint day.
Questions?
Explore the live product at processhq.ai before Sprint Day. The more you understand the four modules, the faster you will move on Sprint morning.