Set up GitHub Copilot properly for any project — in one command.
Without configuration, Copilot doesn't know your stack, your conventions, or your team's patterns. CopilotForge fixes that: it analyzes your codebase, finds the right instructions, agents, and skills for your stack, and installs them.
Copilot works best when it understands your project. That means having:
copilot-instructions.md— teaches Copilot your naming conventions, error handling patterns, testing approach, and architecture- Instructions — stack-specific best practices (TypeScript strict mode, React patterns, pytest conventions…)
- Agents & skills — reusable workflows (code review, conventional commits, test generation…)
- Prompts — quick commands tailored to your workflow
CopilotForge discovers and installs all of these. It ships a curated asset index (no runtime fetching) pre-vetted from github/awesome-copilot, the MCP Registry, and community repos.
Scans your project, scores your current Copilot setup, shows what's missing, and installs what you choose.
🔬 Analyzing your project...
TypeScript | React | Vitest | GitHub Actions | Docker
📊 Current Copilot score: 2/10 (F)
🔍 Found 8 relevant assets:
# │ Type │ Name │ Trust
1 │ instruction │ typescript-strict │ ✅
2 │ instruction │ react-patterns │ ✅
3 │ instruction │ vitest-testing │ ✅
4 │ agent │ code-reviewer │ ✅
5 │ skill │ conventional-commit │ ✅
6 │ mcp-server │ github-mcp-server │ ✅
7 │ prompt │ create-test │ ✅
8 │ instruction │ docker-best-practices │ ✅
Which ones to install? (all / numbers / none)
> all
✅ 8 assets installed. Score: 9/10 (A)
After each install, /forge shows a before/after score comparison and asks if you want to continue improving. Stops automatically at 8+ or when nothing new is found.
Search the curated index by describing what you want, without scanning your project.
/discover EntraID authentication
/discover container deployment with Kubernetes
/discover React testing patterns
Works great when you know what you need but not which asset provides it.
The highest-impact single action. Scans your codebase and generates a copilot-instructions.md tailored to your project:
| Detects | Examples |
|---|---|
| Naming conventions | camelCase vars, kebab-case files, PascalCase types |
| Error handling | try-catch vs Result types, custom error classes |
| Import style | Named vs default, file extensions, sorted imports |
| Testing patterns | Framework, describe-it vs flat, mocking approach |
| Architecture | Layer-based vs feature-based, directory structure |
| Code style | Semicolons, quotes, indentation |
Generate instructions for this project
Step 1 — Clone and build (one time)
git clone https://github.com/g-mercuri/copilot-forge.git ~/copilot-forge
cd ~/copilot-forge/mcp-server && npm install && npm run buildStep 2 — Register as a global MCP server in VS Code
Ctrl+Shift+P → "MCP: Edit User Configuration" → add:
{
"servers": {
"copilot-forge": {
"type": "stdio",
"command": "node",
"args": ["/absolute/path/to/copilot-forge/mcp-server/dist/index.js"]
}
}
}Step 3 — Open any project and run in Copilot Chat (Agent Mode)
/forge
CopilotForge scores your Copilot setup from 0 to 10 and gives a letter grade. Use it to understand gaps and track progress.
| Category | Points |
|---|---|
| copilot-instructions.md | 0–2 |
| Coding standards | 0–1 |
| Error handling | 0–1 |
| Testing practices | 0–1 |
| Language instructions | 0–1 |
| Security/governance | 0–1 |
| Prompts | 0–1 |
| Agents/skills | 0–1 |
| Commit conventions | 0–1 |
apply_org_standards pushes org-wide Copilot standards from your org's .github repo to any project. If no source is configured, it generates sensible defaults from detected codebase patterns.
Every asset in the index carries a trust badge:
| Badge | Level | Sources |
|---|---|---|
| ✅ Verified | High | github/awesome-copilot, microsoft/* |
| Medium | GitHub repos with reasonable activity |
All installs are preview-by-default. Additional protections: URL allowlist, SSRF blocking, content scanning, path traversal prevention, audit log at ~/.copilot-forge/audit.jsonl.
| Tool | Description |
|---|---|
discover_assets |
Analyze project stack + keyword search → ranked asset recommendations |
generate_instructions |
Scan codebase → generate copilot-instructions.md |
install_asset |
Safe batch installer with preview → confirm flow |
score_setup |
Score Copilot setup quality (0–10) with gaps and quick wins |
analyze_project |
Deep project scan (languages, frameworks, dependencies) |
apply_org_standards |
Apply org-wide standards from a .github repo |
TypeScript, JavaScript, Python, Go, Rust, Java, Kotlin, C#/.NET, Ruby, PHP, Swift, React, Next.js, Vue, Angular, Svelte, Django, FastAPI, Spring Boot, ASP.NET, Rails, Laravel, Docker, Kubernetes, Terraform, GitHub Actions, AWS, Azure, GCP, Vitest, Jest, pytest, and more.
.github/
├── agents/copilot-forge.agent.md
├── prompts/
│ ├── forge.prompt.md # /forge command
│ └── discover.prompt.md # /discover command
└── skills/
├── agent-governance/
└── conventional-commit/
mcp-server/src/
├── data/asset-index.ts # Curated asset catalog
├── tools/ # One file per MCP tool
├── security/ # URL validation, content scanning, audit
├── core/ # Scoring engine, git operations
└── analyzers/ # Code pattern detection
cd mcp-server
npm run build # tsc → dist/
npm run test # vitest
npm run lint # eslint- 🐛 Bugs — Open an issue
- 💡 New assets — Suggest additions to the curated index
- 🔧 PRs — Improve tools, add curated assets, fix bugs
- 📖 Docs — Clearer instructions, more examples
Please open an issue before submitting large changes.
🔥 CopilotForge — One command to set up Copilot for any project
Built with ❤️ for the GitHub Copilot community