Pull customer reviews from G2, Capterra, and Trustpilot, cluster by theme, and surface the strongest pull-quotes for marketing use. Marketing teams either ignore reviews or cherry-pick the easy ones. This surfaces every theme honestly so product and marketing share the same truth.

This is the whole blueprint, nothing held back: the model, the tools, the exact prompt, the source on disk, and the setup. Everything you need to run Review Aggregator yourself is on this page, and it deploys in an afternoon.

See it running

Agents like Review Aggregator run behind a metrics and analytics dashboard. Here is that pattern, live: every action stays visible and reviewable. Click into it.

crescendo.51ultron.com/kits/observability

What it does

Review Aggregator handles the work end to end:

Scrape reviews from G2, Capterra, and Trustpilot in one workflow
Cluster by theme with a 5-review minimum to filter noise from signal
Surface verbatim pull-quotes ranked by emotional clarity for marketing
Detect recurring objections so product can address them at the source

The anatomy

An agent is a composition, not a prompt: a model, the tools it is allowed to call, the integrations that give those tools reach, and the rules that keep it honest. Here is Review Aggregator on one card.

Review Aggregatorblueprint

Model: claude-sonnet-4-6
Tools: scrape_reviewscluster_themessurface_quotesdetect_objections
Integrations: Browser UseZod
Guardrails: never edit a quote beyond punctuation and obvious typos.; always include reviewer role + company size when surfacing a quote

The instructions

Every run is governed by this prompt. It is short on purpose: enough to make Review Aggregator reliable, not so much that it can misread itself. Copy it as a starting point for any agent you build.

Review Aggregator

You are a customer voice analyst. You turn raw reviews into themed insights and usable pull-quotes. WORKFLOW: 1. Call scrape_reviews from G2, Capterra, Trustpilot, and any user-provided sources. 2. Call cluster_themes to group reviews by topic (onboarding, support, pricing, integrations). 3. Call surface_quotes to extract verbatim, lightly-cleaned quotes ranked by emotional clarity. 4. Call detect_objections that recur across reviews so product can address them. RULES: - Never edit a quote beyond punctuation and obvious typos. - Always include reviewer role + company size when surfacing a quote. - Theme clusters require at least 5 supporting reviews; smaller signals are noise. - Flag any review that violates the source platform's verification policy.

The source, on disk

Three files carry the whole agent: the definition with its tools, the environment it expects, and the setup. Walk them.

review-aggregator/review-aggregator.ts1.2 KB

1import { agent, tool } from "@agent-sdk"

2import { z } from "zod"

4const scrapeInput = z.object({

5 brand: z.string(),

6 sources: z.array(z.enum(["g2", "capterra", "trustpilot"])).min(1),

7})

9export default agent({

10 model: "claude-sonnet-4-6",

11 permissionMode: "bypassPermissions",

12 maxTurns: 22,

13 systemPrompt: `...`, // see System Prompt section above

14 tools: {

15 scrape_reviews: tool({

16 description: "Pull reviews from G2, Capterra, Trustpilot",

17 inputSchema: scrapeInput,

18 execute: async ({ brand, sources }) => { /* Browser Use */ },

19 }),

20 cluster_themes: tool({

21 description: "Group reviews by topic, min 5 per cluster",

22 inputSchema: z.object({ reviews: z.array(z.any()) }),

23 execute: async ({ reviews }) => { /* clusters */ },

24 }),

25 surface_quotes: tool({

26 description: "Extract verbatim quotes ranked by emotional clarity",

27 inputSchema: z.object({ clusters: z.any() }),

28 execute: async ({ clusters }) => { /* quotes */ },

29 }),

30 detect_objections: tool({

31 description: "Find recurring objections across reviews",

32 inputSchema: z.object({ reviews: z.array(z.any()) }),

33 execute: async ({ reviews }) => { /* objections */ },

34 }),

35 },

36})

The business case

If you run an agency, Review Aggregator is also a product. It sells into product marketing, customer marketing, b2b saas, and the numbers work at very small scale.

$2,300

setup you can charge

$380/mo

monthly retainer

jobs off your plate

Ship it

The whole install is a handful of commands and the keys from the env file. Run the list, then point Review Aggregator at real work.

Deploy checklist

0 of 4

That is the entire blueprint: one model, 4 tools, and a prompt. Built by hand it costs you an afternoon. Inside Ultron it costs you a sentence, because you can build this exact agent in the workforce and watch it run before you commit to anything.

Build Review Aggregator in Ultron

Describe what you want

Start from this blueprint or your own words. Ultron reads the intent and picks the tools.

It builds the agent

The model, the tools, the prompt, the console. Assembled in front of you, nothing to wire.

Watch it run

Test it live in the chat, then put it to work. You review what it does, it does the rest.