דלג לתוכן
Core-Code
…
Core-LibBuild Recipes

What do I need to build it?

Not another list of 500 tools — the exact stack to ship a real thing. Pick a goal and get the recommended stack, MCP servers, UI and starter templates, with cost, complexity and the honest trade-offs.

7 battle-tested stacks — no-code to advanced
Build a CRMBuild a Chat AppBuild a SaaSBuild an AgentBuild a Knowledge BaseBuild a Research AgentBuild a Coding Agent

Build a CRM

A custom CRM your sales team actually wants — no engineers required.

No-code$0–45 / mo to start

Best for: Founders & sales teams who want a tailored CRM live this week — not a Salesforce bill.

Pros

  • Live in a day with zero infrastructure to manage.
  • Generous free tiers across the whole stack.
  • You own a real Postgres database, not a locked SaaS.

Cons

  • The builder is opinionated — deep custom logic hits limits.
  • Heavy customization eventually needs real code.
  • Some lock-in to the no-code builder.

Build a Chat App

Realtime + AI chat with streaming, the way Linear & Vercel ship it.

Low-code$0–25 / mo + LLM usage

Best for: Makers adding live chat or an AI assistant to a product — without reinventing streaming.

Pros

  • Streaming, tool-calls & multi-provider built in.
  • Realtime & vector search on a free tier.
  • Swap models without rewriting the app.

Cons

  • Requires real coding, unlike a no-code builder.
  • Token costs scale with conversation length.
  • You handle rate-limiting & abuse.

Build a SaaS

The standard, scalable stack behind modern subscription products.

Pro-code$0–50 / mo + usage at scale

Best for: Devs launching a real product with auth, billing & a database they fully control.

Pros

  • Battle-tested, hireable stack.
  • Scales from MVP to thousands of users.
  • You own every layer — no platform lock-in.

Cons

  • More wiring than a no-code builder.
  • You own billing edge-cases & security.
  • Costs grow with usage at scale.

Build an Agent

A tool-using, MCP-ready agent on a TypeScript-native framework.

Pro-code$0 infra + LLM usage

Best for: Devs building autonomous, tool-calling agents that do real work end to end.

Pros

  • TypeScript end-to-end — same language as your app.
  • MCP-native: plug in tools instantly.
  • Durable execution survives crashes & retries.

Cons

  • Agent reliability needs evals & guardrails.
  • Token costs add up with long loops.
  • Observability & tracing are on you.

Build a Knowledge Base

Docs + semantic search + grounded AI answers over your own content.

Pro-code$0–50 / mo + embeddings

Best for: Teams turning docs, support & wikis into an answer engine that cites its sources.

Pros

  • Own your data & embeddings end to end.
  • Answers cite real sources — far less hallucination.
  • Great DX with MDX + built-in search.

Cons

  • Ingestion & chunking pipeline to maintain.
  • Freshness needs periodic re-indexing.
  • Embedding costs scale with corpus size.

Build a Research Agent

A deep-research agent that searches the live web and writes cited reports.

AdvancedLLM + search / scrape usage

Best for: Teams building answer engines & analysts that browse, read and synthesize at scale.

Pros

  • Grounded, source-cited answers.
  • Live web — not stale training data.
  • Composable, inspectable multi-step research.

Cons

  • Search & scrape APIs cost per query.
  • Latency grows on deep runs.
  • Needs evals to stay accurate.

Build a Coding Agent

An autonomous coding agent that edits repos in a sandbox — the deep end.

AdvancedLLM usage + sandbox / infra

Best for: Power teams building Devin-style agents & dev tooling with real isolation.

Pros

  • Fully autonomous on real repositories.
  • Sandboxed — agent code can’t touch your machine.
  • Scales to many parallel agents.

Cons

  • Heaviest infrastructure & ops burden.
  • Safety & permissioning are critical.
  • Highest token + compute cost.