Implementation Modes

NirmIQ defines what to build, what could go wrong, and proves it works. Your AI coding tools (Claude, Copilot, Cursor) and IDE define how to build it. Three implementation modes accommodate every team's workflow.

Overview

	Mode A: Mission Control	Mode B: AI-Assisted	Mode C: Fully External
Description	NirmIQ plans, you code externally	NirmIQ generates code suggestions	NirmIQ manages, external tools execute
Code Written In	Your IDE + AI tools	NirmIQ preview, then your IDE	Your IDE
AI Context	Copy to clipboard (rich prompt)	Full prompt assembly by NirmIQ	N/A
Verification	Code Intelligence re-scan	Code Intelligence re-scan	Manual or periodic scan
Git Required	Yes (for verification)	Yes	No
Tier	Basic+	Power (Future)	Basic+
Best For	Teams with existing dev workflow	Maximum AI assistance	Teams using Jira/ADO/GitHub for PM

Mode A: Mission Control (Recommended)

NirmIQ is the "glass cockpit" — it shows what needs doing, assembles rich context for AI coding tools, and verifies what was done. You write code wherever you want.

The Flow

NirmIQ: Requirement reaches "Accepted"
    │
    ├── Work item auto-created (status: todo)
    │
    ├── Developer clicks "Copy to AI"
    │   → Copies rich context: requirement + FMEA risks + knowledge base standards
    │
    ├── Developer pastes into Claude / Copilot / Cursor / any AI tool
    │   → AI has full context: what to build, what could go wrong, project standards
    │
    ├── Developer writes code, reviews, pushes to repo
    │
    └── Back in NirmIQ: Code Intelligence re-scan
        → Detects changes in linked files
        → Work item: todo → in_progress → done
        → Requirement: Accepted → Implemented → Verified

The "Copy to AI" Feature

One click copies a rich, structured prompt that includes everything an AI coding tool needs:

## Implement: REQ-047
Auth module shall implement mutex locking on token refresh
to prevent concurrent refresh race conditions

### Context
- Type: Functional
- Priority: High
- Source: Knowledge Base — "Auth Token Refresh Race Condition"

### Linked Risks (from FMEA)
- RISK-012: Concurrent token refresh causes auth failure (S:8 O:5 D:6 RPN:240)
- RISK-013: Stale token served to subsequent request (S:7 O:4 D:5 RPN:140)

### Project Standards (from Knowledge Base)
- Always use parameterized queries for all database operations
- Auth module must serialize token operations — no concurrent refresh
- All API responses must convert UUID objects to strings

### Existing Code Context
- File: src/api/auth.py (234 lines, 8 functions)
- Last scanned: 2024-02-15, commit: abc123

This is the key differentiator: no standalone AI tool has this context. Claude doesn't know your FMEA items. Copilot doesn't know your knowledge base lessons. NirmIQ assembles it all.

Where to Find It

• Requirement Details tab → "Implementation" section → "Copy to AI" button
• Work Items panel → "Copy to AI" button on each work item

Output Formats

• Markdown (default) — paste directly into any AI chat
• JSON — for programmatic integration with AI APIs or custom tools

Mode B: AI-Assisted Code Generation (Future — Power Tier)

Coming Soon

Mode B is planned for the Power tier. It builds on Mode A by having NirmIQ trigger code generation directly.

NirmIQ uses the same rich context from Mode A but sends it directly to an AI provider, returning suggested code:

NirmIQ: User clicks "Generate Implementation" on work item
    │
    ├── NirmIQ builds rich prompt:
    │   - Requirement text + rationale
    │   - Linked FMEA items (what could go wrong)
    │   - Knowledge base context (lessons, standards)
    │   - Existing code context (from last scan)
    │
    ├── Sends to AI provider (Claude API / Gemini / etc.)
    │
    ├── Returns: suggested code + explanation
    │
    ├── User reviews in NirmIQ (read-only preview)
    │   Options: copy to clipboard, download as file, push to branch
    │
    └── User applies in their IDE

This is NOT another VS Code. It generates a starting point that the developer takes into their real tool. The value is the prompt — NirmIQ knows the requirement, the risks, the standards, and the existing code.

Mode C: Fully External (NirmIQ Manages Process)

NirmIQ is the system of record for requirements, risk analysis, and traceability. Implementation is tracked entirely in external project management tools. No Code Intelligence or Git connection needed.

The Flow

NirmIQ: Requirement reaches "Accepted"
    │
    ├── Work item auto-created in NirmIQ (always — for health score)
    │
    ├── Work item pushed to external tool (user configures):
    │   ├── Jira → Creates issue with [REQ-ID] in title
    │   ├── Azure DevOps → Creates work item (future)
    │   ├── GitHub Issues → Creates issue with labels (future)
    │   └── External ID stored on NirmIQ work item
    │
    ├── Developer works in external tool:
    │   Jira "In Progress" → NirmIQ work item updates (bidirectional sync)
    │   Jira "Done" → NirmIQ work item "done"
    │               → Requirement "Implemented"
    │
    └── Optional: Connect repo later for Code Intelligence verification

External Tool Support

Tool	Status	Sync Direction	Tier
Jira	Available	Bidirectional (push + status sync)	Basic+
Azure DevOps	Planned	Bidirectional	Advanced+
GitHub Issues	Planned	Bidirectional	Advanced+
Linear	Planned	Bidirectional	Advanced+
Webhook	Planned	Push only (you handle sync)	Power+

When to Use Mode C

• Your team already uses Jira/ADO for task management
• You don't have a Git repository (or don't want to connect one)
• You want NirmIQ for requirements/risk/traceability and your existing tool for execution
• Multiple teams contribute — some use NirmIQ directly, others only see Jira tickets

Choosing Your Mode

Most teams use a combination. Mode A for developers who want AI context. Mode C for PMs who track in Jira. They work together:

Requirements in NirmIQ
    ├── Developer: Uses Mode A (Copy to AI → code in IDE)
    └── PM: Uses Mode C (Tracks in Jira → status syncs back)

Both paths feed the same health score and traceability chain.

By Team Type

Team Type	Recommended Mode	Why
AI-first dev team	Mode A	Maximum AI leverage with full context
Enterprise with Jira	Mode A + C	Developers get AI context, PMs track in Jira
No-code/low-code team	Mode C	Requirements and risk management only
Safety-critical engineering	Mode A + C	Full traceability + external PM tool

By Tier

Tier	Available Modes
Free	View requirements, manual status only
Basic	Mode A + Mode C (Jira)
Advanced	Mode A + Mode C (Jira + ADO + GitHub)
Power	All three modes
Enterprise	All three + CI/CD integration + webhooks

What's Next?

• Requirements Lifecycle - Status gates and acceptance flow
• Code Intelligence - AI-powered code scanning
• Jira Integration - Connect Jira for Mode C
• Knowledge Base - Feed lessons into AI suggestions