# AI: tools in this layer of The Stack

> Models, embeddings and agent access. 8 tools, from Nicolas Dao's defaults to what he would never use (Cloudless Labs, Sydney).

- Web page: https://cloudlesslabs.com/stack/layers/ai/

## My default

- [OpenRouter](https://cloudlesslabs.com/stack/openrouter/index.md), LLM access for client systems: One SDK, one key, every model. Usually cheaper than running the same open models elsewhere.
- [pgvector](https://cloudlesslabs.com/stack/pgvector/index.md), Vectors in your app's database: Embeddings next to the data they describe, for when a user is waiting for the answer.

## Use when it fits

- [Amazon S3 Vectors](https://cloudlesslabs.com/stack/aws-s3-vectors/index.md), Vector archive: Vectors at S3 prices, for huge, cold or per-client collections where 100 ms is fine.
- [Cloudflare Sandbox](https://cloudlesslabs.com/stack/cloudflare-sandbox/index.md), Running untrusted and AI-written code: Where an AI product runs code nobody reviewed: an isolated container per session, started and driven from a Worker, billed only for the CPU it actually uses.
- [Self-hosted MCP servers](https://cloudlesslabs.com/stack/mcp-servers/index.md), Agent integrations: Give agents safe, scoped access to client systems. I've started hosting my own.

## Watching

- [Cloudflare AI Gateway](https://cloudlesslabs.com/stack/cloudflare-ai-gateway/index.md) (Proposed), LLM traffic control: Free caching, rate limits, spend caps and model fallback in front of OpenRouter. Works even from Lambda.

## Skip

- [Anthropic API (for client systems)](https://cloudlesslabs.com/stack/anthropic-api/index.md), Claude models, by the token: Too expensive for production features in client systems. OpenAI models give better cost per intelligence. Instead: OpenRouter, picking the model by cost per intelligence.
- [Cloudflare Workers AI](https://cloudlesslabs.com/stack/cloudflare-workers-ai/index.md) (Proposed), Open models on Cloudflare's GPUs: Many of the same open models cost up to twice as much as on OpenRouter, and the newest frontier models are capped at 20 requests a minute. Instead: OpenRouter: the same open models, often for half the price, and far more of them.

## Related

- [An AI chat or LLM feature](https://cloudlesslabs.com/stack/use-cases/ai-chat/index.md): a job that picks or avoids tools from this layer
- [Semantic search and RAG](https://cloudlesslabs.com/stack/use-cases/search/index.md): a job that picks or avoids tools from this layer
- [Building with AI agents](https://cloudlesslabs.com/stack/use-cases/ai-dev/index.md): a job that picks or avoids tools from this layer
- [Every layer, use case and tool in one file](https://cloudlesslabs.com/stack/brief.md)
- [Map of every Markdown file on cloudlesslabs.com, with what each one answers](https://ask.cloudlesslabs.com)