Want AI to do the work?

3 ways to build an AI agent

No-code, developer & open source compared (observed pricing)

Three approaches to building an AI agent — by ease, flexibility and cost. Options by characteristic, not recommendations.

Last updated 2026-06-06 · prices observed 2026-05-30 (official pages)

No-code

Visual builder

Agents without code

For: Build without code

DifyNo-code agent builderFree tier · usage-based

ZapierConnect appsofficial

⚖ Zapier vs Make →

OpenRouterModel routingUsage-based (per token)

Free to start (usage-based / self-host)

+ infra free to start

Save this stack

Developer First

Flexible · controllable

Full control of behavior

For: Engineers who want control

ClaudeReasoning & design$20/month

⚖ Claude vs ChatGPT →

LangGraphAgent orchestrationFree · open source

QdrantVector DB (memory)Free tier · OSS

OpenRouterModel routingUsage-based

$20〜/mo (sum of paid entries, observed)

+ infra free to start

Save this stack

Open Source / Self-host

Low cost · self-hosted

Lowest running cost

For: Keep costs low, own it

n8nWorkflow orchestrationFree · OSS (your server)

OpenRouterModel routingUsage-based

PostgreSQLData & memoryFree · OSS

Free to start (usage-based / self-host)

+ infra free to start

Save this stack

Representative patterns split by speed / flexibility / cost. No subjective claims or adoption rates (adoption data not yet accumulated).

Observation only — no predictions, no recommendations. Representative stacks shown with observed pricing and sources. You decide what to build. © TraceBottle for AI