Starter stack

An AI agent that does real work

Your product is an agent — it plans, calls tools and APIs, waits on people or other systems, and runs for minutes or hours on a user's behalf. Reliability and cost matter more than a chat box.

25 makers' products and 21 open-source projects use at least half of these 8 tools.

What it costs

Launchabout 1,000 users: $26/mo
Tractionabout 10,000 users: $203/mo
Scaleabout 100,000 users: $509/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: An AI agent that does real work

Your product is an agent — it plans, calls tools and APIs, waits on people or other systems, and runs for minutes or hours on a user's behalf. Reliability and cost matter more than a chat box.

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

- Frontend Framework: Next.js — https://nextjs.org
agent docs: https://nextjs.org/llms.txt · CLI: npx next
Why: The app where users set goals, approve steps and watch runs.
- AI SDK & Agent Framework: Mastra — https://mastra.ai
agent docs: https://mastra.ai/llms.txt · CLI: npx mastra
Why: A TypeScript agent framework with workflows, memory and tool calling, built on the AI SDK.
- LLM API: Claude — https://www.anthropic.com/claude
agent docs: https://docs.anthropic.com/llms.txt
Why: Strong at long, multi-step tool use — the core of an agent that has to finish the job.
- Background Jobs & Cron: Inngest — https://www.inngest.com
agent docs: https://www.inngest.com/llms.txt · CLI: npx inngest-cli
Why: Durable steps, so a run survives restarts, retries a failed call without redoing the rest, and can wait days for a human approval.
- Database: Supabase — https://supabase.com
agent docs: https://supabase.com/llms.txt · MCP server: com.supabase/mcp (https://mcp.supabase.com/mcp) · CLI: npx supabase
Why: Users, agent state and run history in Postgres, with realtime updates to show progress live.
- LLM Observability & Evals: Langfuse — https://langfuse.com
agent docs: https://langfuse.com/llms.txt · CLI: npx langfuse-cli
Why: Every step, tool call and token in a run, so you can debug a bad outcome and see what each run costs.
- Payments: Autumn — https://useautumn.com
agent docs: https://useautumn.com/llms.txt · CLI: npx atmn
Why: Agent runs cost real money in tokens; credits and limits keep heavy users from sinking your margin.
- Hosting: Vercel — https://vercel.com
agent docs: https://vercel.com/llms.txt · MCP server: com.vercel/vercel-mcp (https://mcp.vercel.com) · CLI: npx vercel
Why: Serverless for the app, with long work handed off to the job runner.

Trade-offs: Agents fail in new ways — loops, runaway costs, confident wrong actions. Build limits, approvals and evals in from the first version. Python-first teams may prefer LangGraph or Pydantic AI with the same shape.

Source: https://howmakersbuild.com/stacks/ai-agent-product

The picks

Frontend Framework

Next.jsllms.txt CLI

The app where users set goals, approve steps and watch runs.

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

AI SDK & Agent Framework

Mastrallms.txt CLI

A TypeScript agent framework with workflows, memory and tool calling, built on the AI SDK.

Chosen for AI SDK & agent framework by 24 makers' products we track · the alternatives →

LLM API

Claudellms.txtIn Lovable, Replit

Strong at long, multi-step tool use — the core of an agent that has to finish the job.

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

Background Jobs & Cron

Inngestllms.txt CLIIn Lovable

Durable steps, so a run survives restarts, retries a failed call without redoing the rest, and can wait days for a human approval.

Chosen for background jobs & cron by 16 makers' products we track · the alternatives →

Database

Supabasellms.txt MCP CLIIn Bolt, Lovable, v0

Users, agent state and run history in Postgres, with realtime updates to show progress live.

Chosen for database by 564 makers' products we track · the alternatives →

LLM Observability & Evals

Langfusellms.txt CLI

Every step, tool call and token in a run, so you can debug a bad outcome and see what each run costs.

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

Payments

Autumnllms.txt CLI

Agent runs cost real money in tokens; credits and limits keep heavy users from sinking your margin.

Chosen for payments by 12 makers' products we track · the alternatives →

Hosting

Vercelllms.txt MCP CLIIn v0

Serverless for the app, with long work handed off to the job runner.

No maker's product we track shows it for hosting yet · the alternatives →

What this costs you

Agents fail in new ways — loops, runaway costs, confident wrong actions. Build limits, approvals and evals in from the first version. Python-first teams may prefer LangGraph or Pydantic AI with the same shape.

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.