Starter stack

Python AI backend with a web front end

The AI part of your product lives in Python — models, data work, the libraries researchers publish first — and the front end is a separate web app talking to it.

0 makers' products and 35 open-source projects use at least half of these 9 tools.

What it costs

Launchabout 1,000 users: $27/mo
Tractionabout 10,000 users: $181/mo
Scaleabout 100,000 users: $1,267/mo

Monthly bills of the priced picks, list prices.

Do the picks fit together?No conflicts: no acquired or archived tool, no two tools for a decision that usually has one, nothing the plan's rules say doesn't fit.Open in Check my stack — add or swap tools →

Hand it to your AI agent

Paste into AGENTS.md, CLAUDE.md or your Cursor rules: each tool with its agent docs, MCP server and CLI where they exist.

Preview

# Stack: Python AI backend with a web front end

The AI part of your product lives in Python — models, data work, the libraries researchers publish first — and the front end is a separate web app talking to it.

Use these tools. Ask before swapping one for an alternative.

- Backend Framework: FastAPI — https://fastapi.tiangolo.com
  Why: Typed request and response models, async handlers for slow model calls, and automatic OpenAPI docs your front end can generate a client from.
- Database Access & ORMs: SQLAlchemy — https://www.sqlalchemy.org
  Why: The standard Python data layer, with Alembic for migrations.
- Database: PostgreSQL — https://www.postgresql.org
  Why: One database for app data and, with pgvector, embeddings too.
- AI SDK & Agent Framework: Pydantic AI — https://ai.pydantic.dev
  agent docs: https://ai.pydantic.dev/llms.txt
  Why: A small, typed agent loop in the same Pydantic style as FastAPI — structured outputs you can validate.
- LLM API: Claude — https://www.anthropic.com/claude
  agent docs: https://docs.anthropic.com/llms.txt
  Why: Strong at long documents and tool use; switch per task when another model is cheaper.
- Vector Database: Qdrant — https://qdrant.tech
  agent docs: https://qdrant.tech/llms.txt
  Why: An open-source vector store with filtering and hybrid search, self-hosted or managed.
- LLM Observability & Evals: Langfuse — https://langfuse.com
  agent docs: https://langfuse.com/llms.txt · CLI: npx langfuse-cli
  Why: Traces, costs and evals with a first-class Python SDK, self-hostable.
- Frontend Framework: Next.js — https://nextjs.org
  agent docs: https://nextjs.org/llms.txt · CLI: npx next
  Why: The web app in front, calling the FastAPI backend through a generated, typed client.
- Hosting: Render — https://render.com
  agent docs: https://render.com/llms.txt
  Why: Runs a long-lived Python server, background workers and Postgres side by side, without servers to manage.

Trade-offs: Two languages mean two deploys, two sets of dependencies and an API contract to keep in sync. If the AI work is mostly calling model APIs, a TypeScript-only stack is simpler; go Python when you need its libraries.

Source: https://howmakersbuild.com/stacks/python-ai-backend

The picks

Backend Framework

FastAPI

Typed request and response models, async handlers for slow model calls, and automatic OpenAPI docs your front end can generate a client from.

Chosen for backend framework by 29 makers' products we track · the alternatives →

Database Access & ORMs

SQLAlchemy

The standard Python data layer, with Alembic for migrations.

No maker's product we track shows it for database access & ORMs yet · the alternatives →

Database

PostgreSQLIn Replit

One database for app data and, with pgvector, embeddings too.

Chosen for database by 74 makers' products we track · the alternatives →

AI SDK & Agent Framework

Pydantic AIllms.txt

A small, typed agent loop in the same Pydantic style as FastAPI — structured outputs you can validate.

No maker's product we track shows it for AI SDK & agent framework yet · the alternatives →

LLM API

Claudellms.txtIn Lovable, Replit

Strong at long documents and tool use; switch per task when another model is cheaper.

Chosen for LLM API by 251 makers' products we track · the alternatives →

Vector Database

Qdrantllms.txt

An open-source vector store with filtering and hybrid search, self-hosted or managed.

Chosen for vector database by 18 makers' products we track · the alternatives →

LLM Observability & Evals

Langfusellms.txt CLI

Traces, costs and evals with a first-class Python SDK, self-hostable.

Chosen for LLM observability & evals by 33 makers' products we track · the alternatives →

Frontend Framework

Next.jsllms.txt CLI

The web app in front, calling the FastAPI backend through a generated, typed client.

Chosen for frontend framework by 1510 makers' products we track · the alternatives →

Hosting

Renderllms.txt

Runs a long-lived Python server, background workers and Postgres side by side, without servers to manage.

Chosen for hosting by 128 makers' products we track · the alternatives →

What this costs you

Two languages mean two deploys, two sets of dependencies and an API contract to keep in sync. If the AI work is mostly calling model APIs, a TypeScript-only stack is simpler; go Python when you need its libraries.

Ask your AI about this, with this page as the source:ChatGPT ↗Claude ↗Perplexity ↗

Who builds this way

Using at least half of these tools, each with a source: makers' products from their own words, customer stories, subprocessor lists or websites; open-source projects from their code.

awesome-llm-appsopen source100+ AI Agents, Agent Skills and RAG Apps - Free and Open Source.7 of these 9 tools AutoGPTopen sourceAutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.6 of these 9 tools camelopen source🐫 CAMEL: The first and the best multi-agent framework. Finding the Scaling Law of Agents. https://www.camel-ai.org6 of these 9 tools FinSynthopen source基于Open WebUI框架的金融大模型交互平台——支持多模态非结构化数据处理6 of these 9 tools puppyone-cloudopen sourceContext drive for your AI agents6 of these 9 tools agentfieldopen sourceBuild, run and scale AI agents like API and microservices5 of these 9 tools ever-worksopen sourceEver® Works™ - The Workshop for AI. An open agentic runtime that autonomously researches, ships, and maintains entire businesses, 24/7 - https://ever.works5 of these 9 tools monitoropen sourceReal-time monitoring, slowlog analysis, and audit trails for Valkey and Redis5 of these 9 tools

Other starter stacks

Ship an AI SaaS this weekend Sell worldwide without handling tax AI app that bills by usage Build it with an AI app builder Chat with your docs (RAG)B2B SaaS that sells to companies $0 until you have users An AI agent that does real work The one-person monolith Edge-first on Cloudflare