A tool-agnostic template catalog for AI-focused projects. Each template is a self-contained, scaffoldable starting point — consumable by any scaffolding tool that reads the template contract.
New here? Start with the Quick Start guide — find the right template, scaffold it, and run it in under 2 minutes.
48 templates across four categories, plus 15 composable modules.
| Template | Description | Tags |
|---|---|---|
| agentic-loop | Autonomous agent with tool registry, memory, provider abstraction, eval harness | typescript agent llm |
| rag-pipeline | TypeScript RAG pipeline with chunking, embedding, vector store, retrieval, generation | typescript rag embeddings |
| mcp-server-ts | TypeScript MCP server with tool/resource registration and Zod validation | mcp typescript tools |
| mcp-server-python | Python MCP server with tool/resource registration | mcp python tools |
| eval-harness | Standalone LLM evaluation framework with YAML suites, assertions, and reporters | typescript eval testing |
| prompt-library | Versioned prompt management with YAML frontmatter and optional TypeScript SDK | prompts yaml versioning |
| a2a-agent | Agent-to-Agent protocol peer with skill registry, transport abstraction, and discovery | typescript a2a agents |
| guardrails | Input/output safety filters — prompt injection, PII, content policy, token limits | typescript safety filters |
| multimodal-pipeline | Process images, audio, and video with AI models | typescript multimodal ai |
| fine-tune-pipeline | LLM fine-tuning pipeline with dataset preparation and training | typescript fine-tuning llm |
| data-pipeline | Data ingestion and transformation pipeline | typescript data pipeline |
| agent-orchestrator | Multi-agent orchestration with delegation and coordination | typescript agents orchestrator |
| ci-eval | CI-integrated LLM evaluation with GitHub Actions | typescript eval ci |
| Template | Description | Tags |
|---|---|---|
| go-cli | Cobra CLI with Viper config, slog logging, GoReleaser release workflow | go cli cobra |
| ts-service | Hono HTTP service with database drivers, JWT auth, and OpenTelemetry | typescript hono api |
| go-service | Go HTTP service with chi router, repository pattern, and OpenTelemetry | go chi api |
| chrome-ext | Chrome extension (Manifest V3) with React sidepanel for AI interactions | chrome-extension react vite |
| vscode-ext | VS Code extension with optional React webview and AI provider integration | vscode extension typescript |
| next-app | Next.js 15 App Router with streaming AI chat, Tailwind, auth, and database | typescript nextjs react |
| slack-bot | Slack bot with event handling and slash commands | typescript slack bot |
| discord-bot | Discord bot with command registration and event handling | typescript discord bot |
| electron-app | Electron desktop application with IPC and auto-updates | typescript electron desktop |
| api-gateway | API gateway with routing, rate limiting, and upstream management | typescript gateway api |
| worker-service | Background worker service with job processing | typescript worker jobs |
| Template | Description | Tags |
|---|---|---|
| infra-aws | AWS CDK with Lambda/ECS, VPC, RDS, and CloudWatch monitoring | aws cdk infrastructure |
| infra-fly | Fly.io deployment with Dockerfile, fly.toml, and CI/CD | fly deployment docker |
| infra-gcp | GCP Cloud Run with Cloud SQL, monitoring, and GitHub Actions | gcp cloud-run infrastructure |
| infra-vercel | Vercel deployment with edge functions and CI/CD | vercel deployment edge |
| infra-druid | Apache Druid cluster with ingestion and query configuration | druid analytics infrastructure |
| infra-cloudflare | Cloudflare Workers with KV, D1, and R2 | cloudflare workers infrastructure |
| k8s-deploy | Kubernetes manifests and Helm chart with Ingress, HPA, and CI/CD | kubernetes helm deployment |
| monorepo | Turborepo/pnpm workspace with shared packages and CI | monorepo turborepo pnpm |
| monitoring-stack | Observability infrastructure with metrics, logs, and traces | monitoring observability infrastructure |
| Template | Description | Tags |
|---|---|---|
| module-auth-ts | Authentication middleware with pluggable providers (JWT, Clerk, Auth0, Supabase, API key) | auth middleware jwt |
| module-database-ts | Drizzle ORM with pluggable drivers (PostgreSQL, SQLite, Turso) | database drizzle orm |
| module-observability-ts | OpenTelemetry instrumentation with pluggable exporters (console, OTLP, Datadog) | observability opentelemetry tracing |
| module-storage-ts | Blob storage abstraction with pluggable backends (local, S3, R2, GCS) | storage s3 blob |
| module-queue-ts | Background job processing with pluggable brokers (in-memory, BullMQ, SQS) | queue jobs workers |
| module-cache-ts | Caching layer with pluggable backends (memory, Redis) | cache redis performance |
| module-rate-limit-ts | Rate limiting with pluggable algorithms (token bucket, sliding window) | rate-limit throttle middleware |
| module-webhook-ts | Webhook ingestion and delivery with signature verification | webhook events middleware |
| module-notifications-ts | Multi-channel notifications (email, SMS, push) | notifications email messaging |
| module-llm-gateway | LLM provider gateway with routing and fallback | llm gateway ai |
| module-vector-store | Vector store abstraction with pluggable backends | vectors embeddings ai |
| module-semantic-cache | Semantic caching for LLM responses using embeddings | cache semantic ai |
| module-llm-observability | LLM-specific observability with cost and latency tracking | observability llm cost |
| module-billing-ts | Usage-based billing and subscription management | billing payments saas |
| module-feature-flags-ts | Feature flag management with evaluation rules | feature-flags toggles configuration |
Every template lives in templates/<name>/ and contains:
template.yaml— declares metadata, variables, conditionals, and hooks per the template contract specskeleton/— the file tree to scaffold, with placeholder strings for variable interpolationREADME.md— when and why to reach for this template
Scaffolding tools read template.yaml, collect variable values from the user, and render skeleton/ with those values substituted for their declared placeholders. The contract is generic — no tool-specific coupling.
Pre-configured multi-template stacks in composites/:
| Composite | Use case |
|---|---|
| proof-of-concept | Minimal AI agent + evals — fastest path to a working prototype |
| ai-chatbot | Customer-facing chatbot with agent, service, auth, and deployment |
| document-intelligence | Document search and Q&A with RAG, storage, and database |
| rag-agent | AI agent using RAG as a knowledge tool |
| safe-ai-agent | Agent with guardrails, evals, and prompt management |
| multi-agent | A2A protocol peers with MCP tools, guardrails, and orchestration |
| ai-web-app | Full-stack Next.js app with RAG, auth, database, and observability |
| mcp-toolkit | MCP server with prompts and evals |
| chrome-ai-extension | Chrome extension backed by MCP tools |
| vscode-ai-extension | VS Code extension with AI, MCP, and prompts |
| production-api | Full-stack API with auth, database, cache, rate limiting, queue, observability |
| go-microservice | Go service with evals and Fly.io deployment |
| background-processor | Async job processing with queue, storage, and database |
| internal-tool | CLI or browser extension with MCP tools |
| enterprise-ai | Full enterprise stack — agents, RAG, guardrails, evals, K8s |
| eval-suite | Standalone eval infrastructure with prompts and safety testing |
| cost-optimized-ai | LLM gateway with cost-aware routing, semantic caching, and cost tracking |
| ai-platform | Full AI platform with service, gateway, vectors, pipeline, auth, billing, monitoring |
| agent-team | Multi-agent system with orchestrator, specialized agents, evals, and MCP tools |
Composites define which templates to scaffold, where to nest them, and how variables flow across templates. See the composite contract for the specification.
@nanohype/sdk is the reference implementation of the template rendering contract. Install it as a dependency to build your own scaffolding tool, CLI, MCP server, or CI pipeline on top of nanohype templates.
import { LocalSource, renderTemplate } from '@nanohype/sdk';
const source = new LocalSource({ rootDir: './nanohype' });
const { manifest, files } = await source.fetchTemplate('agentic-loop');
const result = renderTemplate(manifest, files, { ProjectName: 'my-agent' });
// result.files — rendered SkeletonFile[] ready to write to diskThe SDK lives at sdk/ in this repo. See the Consumer Implementation Guide for the full scaffolding algorithm.
- Consumer Implementation Guide — scaffolding algorithm, composite orchestration, catalog integration
- Catalog Reference — decision matrix, composition map, and template selection guide
- Template Contract — the formal schema for
template.yaml - Composite Contract — multi-template stack manifests
- Authoring Guide — practical walkthrough for creating new templates
- Contributing — how to add a template to the catalog
npm install
npm run validate:schemaOr validate a single template:
./scripts/validate.sh templates/go-cli