Skip to main content

CoPilot
Collections

Stop switching between Jira, Figma, and your codebase. One framework. Every phase of delivery. AI at every step.

See on GitHubRead the docs

Built by The Software House · Used daily by 300+ engineers · 50+ commercial projects

A complete AI
product engineering
framework — not
just a set of prompts

End-to-End Product Lifecycle

Covers Product Ideation, Development, and Quality in a single framework. From workshop transcript to production-ready code — every phase has dedicated agents, skills, and workflows.

10 Specialized AI Agents

Business Analyst, Architect, Software Engineer, Code Reviewer, UI Reviewer, E2E Engineer, DevOps Engineer, Context Engineer, Copilot Engineer, and Copilot Orchestrator — each focused on its phase, working together in a structured sequence.

Structured Delivery Workflow

Research → Plan → Implement → Review. Each phase feeds the next. Research becomes the plan. The plan becomes the implementation checklist. Review returns issues to the exact phase.

Pixel-Perfect UI Verification

Automated Figma comparison loop in every UI implementation cycle. Playwright captures the rendered state and compares it to the design spec. Mismatches are fixed before review.

Requirements Processing

Turn workshop transcripts and meeting notes into Jira-ready user stories before a line of code is written. The framework starts at the source of the work, not just the ticket.

MCP Tool Integrations

Jira, Figma Dev Mode, Context7, Playwright, Sequential Thinking, AWS API, GCP Gcloud, and more — wired into the workflow phases, not bolted on. Context flows through every step automatically.

Used daily by 300+
engineers at TSH

Read the methodology →
01

Average lead time reduction: 30% — measured across 50+ commercial projects.

02

From juniors to principals. Full adoption across all seniority levels.

This is the
new SDLC.

Every phase of delivery restructured around how AI actually works — not bolted onto an existing process. From raw workshop transcripts to production-ready code.

Every step requires your review. The framework provides structure — judgment stays with your team.

Core use cases

Every use case has dedicated agents, prompts, and workflows — not generic AI assistance.

Product Ideation
Workshop Outputs to Jira Backlog
Discovery workshops produce valuable transcripts, Figma boards, and shared notes — but converting them into structured, actionable Jira tickets is a manual, error-prone process. Tasks are vague, edge cases are missed, and the backlog doesn’t reflect what was actually discussed.
~15 min vs 1–2 days
Development
Single-Session Context Gathering
Requirements live in Jira, designs in Figma, documentation in Confluence, code in GitHub. Developers constantly context-switch to gather information. MCP integrations bring all context into a single Copilot chat session.
~3 min vs 30–60 min
Quality
Automated Code Review
Multi-dimensional review checking acceptance criteria, security, performance, database patterns, and coding standards. Clear PASS/BLOCKER/SUGGESTION verdicts on every run.
~5 min
Development
Pixel-Perfect UI Delivery
Frontend implementations deviate from Figma — wrong spacing, colours, component variants. QA catches these late, causing rework. The automated Figma verification loop runs up to 5 iterations, comparing the running app via Playwright against Figma specs before the code ever reaches human review.
95–99% design accuracy
Quality
Security Built Into Every Plan and Review
Security reviews happen at the end of a sprint — if at all. DRY, KISS, and proper error handling are inconsistently applied. Every plan includes security considerations by default. The Code Reviewer agent checks for vulnerabilities, missing input validation, and exposed secrets on every run.
Built in, every time
Development
Onboarding New Team Members
New developers spend days understanding the codebase, conventions, and task requirements before they can contribute. The /tsh-research prompt gathers context from Jira, Confluence, Figma, and the codebase automatically. /tsh-plan creates a step-by-step implementation path — within minutes, not days.
~5 min per task
Development
Architecture Plan from a Ticket
Generates a phased implementation plan with CREATE/MODIFY/REUSE labels, security considerations, and clear definitions of done — directly from a Jira ticket.
~5 min
Quality
Reliable E2E Test Suites
E2E tests are written inconsistently, use brittle selectors, and break on unrelated changes. Teams stop trusting them and start skipping them. The E2E Engineer agent enforces Page Object patterns, accessibility-first locators, and verifies tests for 3+ consecutive passes before commit.
~10 min per test suite
Infrastructure
Cloud Cost Optimization
Cloud bills keep growing but nobody knows which resources are wasteful. The DevOps Engineer agent performs hybrid audits — analyzing IaC code and validating against live infrastructure via AWS/GCP APIs to identify savings opportunities.
~10 min per audit

Set up once.
Works across
every project.

Clone the repo next to your projects, configure VS Code User Settings once, and start using /research in any workspace immediately.

Get the repo on GitHubRead the docs
1

Clone the repo

git clone [repo] copilot-configuration alongside your existing projects

2

Configure VS Code

Open User Settings (JSON) → add chat.promptFilesLocations and chat.modeFilesLocations pointing to the cloned folder. Done once — works globally.

3

Configure MCP servers

Copy .vscode/mcp.json to your User MCP config. Connects Jira, Figma, Playwright, and Context7.

4

Run your first command

Open Copilot Chat → select an agent → type /research [JIRA_ID]. If agents appear in the dropdown, you're ready.