{"id":581,"date":"2026-10-08T10:06:00","date_gmt":"2026-10-08T10:06:00","guid":{"rendered":"https:\/\/katteb.com\/blog\/?p=581"},"modified":"2026-10-08T10:06:00","modified_gmt":"2026-10-08T10:06:00","slug":"perplexity-ai-agent","status":"publish","type":"post","link":"https:\/\/katteb.com\/blog\/perplexity-ai-agent\/","title":{"rendered":"Perplexity AI Agent: Use Cases, Perplexity API Pricing and What a Research Agent Really Costs"},"content":{"rendered":"<div style=\"border-left:5px solid #7b61ff;background:#f6f4ff;padding:16px 20px;margin:0 0 24px\">\n<p style=\"margin:0\"><strong>TL;DR.<\/strong> A Perplexity AI agent is a system that uses Perplexity&#8217;s search-grounded answers to do recurring research work: competitor briefs, market questions, account research, pricing watches and cited summaries. You can build one on the official Perplexity API, which is billed per token and per tool call (a light research run costs about half a cent in Perplexity&#8217;s own example), or run the same research as a goal with a browser-based agent such as Katteb, which is a flat $19 or $99 a month. On raw API spend, Perplexity is usually the cheaper line item for light workloads; what you pay for with an agent platform is the scheduling, memory, delivery and the ability to act in your own accounts. This guide covers the use cases, the official prices as published on 8 October 2026, three worked cost scenarios with every assumption stated, and a way to decide.<\/p>\n<\/div>\n<p>Perplexity became popular because it answers a question with sources attached. For a lot of teams that is the first piece of an AI agent: something that can look up what is true right now and show where it found it. The second piece is everything around the answer. Who asks the question every Monday? Where does the result go? Does anyone compare this week&#8217;s answer with last week&#8217;s? Does anything happen after the answer arrives?<\/p>\n<p>That is why people search for &#8220;Perplexity API pricing&#8221; before they build. They are trying to work out what a research agent will cost to run, and whether they should wire one up themselves or use a product that already does the surrounding work. This article answers both questions with numbers you can check.<\/p>\n<h2>What is a Perplexity AI agent?<\/h2>\n<p>A Perplexity AI agent is an AI system that repeatedly asks the web questions through Perplexity, reads the cited sources, and turns the results into something useful without a person driving each step. In practice it combines four parts:<\/p>\n<ul>\n<li><strong>A search-grounded model<\/strong>, such as Perplexity&#8217;s Sonar model or a third-party model called through Perplexity&#8217;s Agent API with web search switched on.<\/li>\n<li><strong>Tools<\/strong>: web search, fetching a specific URL, people search, finance search, or a code sandbox for calculations.<\/li>\n<li><strong>A plan and a schedule<\/strong>: which questions to ask, how often, and what counts as a change worth reporting.<\/li>\n<li><strong>A destination<\/strong>: an email, a Slack message, a spreadsheet row, a CRM note, a report.<\/li>\n<\/ul>\n<p>The first two parts are what Perplexity sells by the call. The last two are what you either build yourself or get from an agent platform. Keeping that split in mind makes the pricing question much easier.<\/p>\n<h2>Which Perplexity surface are people actually talking about?<\/h2>\n<p>&#8220;Perplexity&#8221; can mean two quite different things, and mixing them up leads to wrong cost estimates.<\/p>\n<h3>The developer API<\/h3>\n<p>Perplexity&#8217;s API platform is aimed at developers. Its documentation lists an Agent API (one interface to models from Perplexity, OpenAI, Anthropic, Google, xAI and others, with built-in tools), a Search API that returns raw web results, an Embeddings API, a Router API and a Decisions API. You pay for what you use: tokens in, tokens out, and each tool invocation.<\/p>\n<h3>The consumer and team apps<\/h3>\n<p>Perplexity Pro and Max are subscriptions for people using the app. On its plans page Pro is listed at $17 a month and Max at $167 a month, both &#8220;or equivalent, when billed annually&#8221;, and both marked &#8220;for personal use only&#8221;. They include Perplexity&#8217;s own agent feature, Computer, plus bonus credits. These are tools for a person doing research, not a metered backend for software you build.<\/p>\n<h3>Why the difference matters<\/h3>\n<p>If you are building a product or an automated pipeline, the API is the relevant price list. If you want a researcher to have better answers, the app subscription is. If you want research work to happen on a schedule and lead to actions, you are comparing the API (plus your own engineering) against an agent platform. The rest of this article is mostly about that third case.<\/p>\n<h2>What does the Perplexity API cost?<\/h2>\n<p>The figures below were read from Perplexity&#8217;s official pricing and model pages on 8 October 2026. All prices are in US dollars, before tax. Perplexity can change them at any time, so check the live pages before you budget.<\/p>\n<h3>Agent API tools<\/h3>\n<table style=\"border-collapse:collapse;width:100%;margin:12px 0 24px\">\n<thead>\n<tr>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Tool<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Price<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Per 1,000<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">web_search (standard)<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.0025 per invocation<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$2.50<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">web_search (Fast Search)<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.001 per invocation<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$1.00<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">fetch_url<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.0005 per invocation<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.50<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">people_search<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.005 per invocation<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$5.00<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">finance_search<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.005 per invocation<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$5.00<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">sandbox<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.03 per session (20-minute billing window)<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">n\/a<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>Tool costs are separate from model token costs, so a single research request is the sum of both.<\/p>\n<h3>Model tokens (selected models)<\/h3>\n<table style=\"border-collapse:collapse;width:100%;margin:12px 0 24px\">\n<thead>\n<tr>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Model in the Agent API<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Input per 1M tokens<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Output per 1M tokens<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">perplexity\/sonar<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.25<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$2.50<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">anthropic\/claude-sonnet-5-5<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$2.00<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$10.00<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">openai\/gpt-5.6-terra (up to 272k context)<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$2.00<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$12.00<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">google\/gemini-3.1-flash-lite<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.25<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$1.50<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>The docs also list a <em>flex<\/em> service tier at half the listed token price (best-effort capacity) and a <em>priority<\/em> tier at twice the price. Reasoning tokens are billed at the output rate, which matters if you turn reasoning effort up.<\/p>\n<h3>Search API and presets<\/h3>\n<p>The Search API, which returns raw results rather than a written answer, is listed at $5.00 per 1,000 successful requests, or $1.00 per 1,000 with Fast Search. A request may contain up to five queries and still counts as one billing unit. Failed and rate-limited requests are not billed.<\/p>\n<p>Perplexity also publishes presets (fast, low, medium, high, xhigh) that bundle a model, search settings and a step budget. Its pricing page gives a worked example for a representative &#8220;low&#8221; preset run of 2,000 input and 1,000 output tokens with one Fast Search and one fetch: $0.0055 in total. Deeper presets run more steps and cost more; Perplexity does not publish a flat per-run price for them.<\/p>\n<h2>What can a Perplexity AI agent do for a business?<\/h2>\n<p>Search-grounded answers are useful wherever a team keeps asking the same kind of question about a world that changes. These are the use cases that come up most.<\/p>\n<h3>1. Weekly competitor briefs<\/h3>\n<p>Ask the same set of questions about five to twenty competitors every week: new pricing, new features, announcements, hiring signals, press coverage. The value is not one answer; it is the comparison with last week, so the brief only flags what changed.<\/p>\n<h3>2. Account research before sales calls<\/h3>\n<p>Before a call, pull a short cited brief on the company: what it sells, recent news, funding, leadership changes, and likely priorities. Perplexity&#8217;s people search tool adds the ability to look up professionals, which is billed separately.<\/p>\n<h3>3. Market and category questions<\/h3>\n<p>&#8220;How big is this market, who leads it, what changed this quarter?&#8221; Questions like these need several sources chained together, which is exactly what the deeper presets are for. Every claim should come with a citation a human can open.<\/p>\n<h3>4. Pricing and offer watch<\/h3>\n<p>Check a list of vendor or competitor pricing pages on a schedule and report differences. A fetch of a specific URL costs a fraction of a cent through the API, but you still need to store the previous version to know what changed.<\/p>\n<h3>5. Content research with sources<\/h3>\n<p>Writers and SEO teams use search-grounded models to gather current facts, statistics and quotes with links, so the draft is checkable. The agent&#8217;s job is to produce a research pack, not the final article.<\/p>\n<h3>6. Finance and earnings summaries<\/h3>\n<p>Perplexity&#8217;s finance search tool returns market and financial data. A scheduled agent can summarise each earnings release for a watchlist and flag numbers outside an expected range.<\/p>\n<h3>7. Vendor and partner due diligence<\/h3>\n<p>Before signing with a supplier, gather reviews, news, legal mentions and ownership details into one cited file. This is a one-off task rather than a schedule, but it follows the same pattern.<\/p>\n<h2>Where does the API stop and the agent begin?<\/h2>\n<p>A Perplexity API call returns an answer with citations. That is a strong building block, and it is where the API&#8217;s job ends. Everything below is still yours to build if you go the API route:<\/p>\n<ul>\n<li><strong>Scheduling<\/strong>: something has to run the questions every Monday at 9am, retry on failure and alert you when it breaks.<\/li>\n<li><strong>Memory<\/strong>: to report only what changed, you need to store previous answers and compare them.<\/li>\n<li><strong>Delivery<\/strong>: an email, a Slack message, a Google Sheet row. Each is another integration.<\/li>\n<li><strong>Actions<\/strong>: if the research should lead to something (update a CRM record, draft a reply, post an update, fill a form), the API does not do it. You write that code.<\/li>\n<li><strong>Signed-in sources<\/strong>: an API searches the public web. It does not read your own dashboards, inboxes or the parts of sites that need a login.<\/li>\n<\/ul>\n<p>An agent platform bundles those pieces. Katteb, for example, runs an agent in a real browser that can work inside the sites you sign it into, remembers what it did, runs on a schedule and reports every day. The trade-off is that you pay a flat subscription instead of fractions of a cent per call, and you get less low-level control than with your own code.<\/p>\n<h2>Perplexity API vs an agent platform: three cost scenarios<\/h2>\n<p>Here is the arithmetic, with every assumption stated so you can swap in your own numbers.<\/p>\n<p><strong>Common assumptions:<\/strong> US dollars; one calendar month (a year is 52 weeks, a month is 52 \u00f7 12 \u2248 4.33 weeks, and 21.67 working days); list prices read on 8 October 2026, before tax; no free credits, discounts or enterprise terms; engineering time, hosting and the cost of the scheduler, database and integrations around the API are <em>not<\/em> included, because they vary too much to estimate honestly. Katteb prices are from katteb.com\/pricing: Lite $19 a month or $190 a year (1 agent, up to 30 hours of agent time a week); Goals $99 a month or $990 a year (10 agents, up to 120 hours a week).<\/p>\n<h3>Scenario A: a light weekly research brief<\/h3>\n<p>25 research questions a week, each run as one &#8220;low&#8221; preset request at Perplexity&#8217;s published example cost of $0.0055.<\/p>\n<table style=\"border-collapse:collapse;width:100%;margin:12px 0 24px\">\n<thead>\n<tr>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Line<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Calculation<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Monthly<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Runs per month<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">25 \u00d7 4.33<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">\u2248 108 runs<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Perplexity API spend<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">108 \u00d7 $0.0055<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">\u2248 $0.60<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Katteb Lite<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">flat<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$19.00<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><strong>Result:<\/strong> on raw spend the Perplexity API is about $18.40 a month cheaper. If you already have a developer and a place to run scheduled jobs, building on the API is the economical route for this workload. Katteb only makes sense here if you would rather not build and maintain the scheduler, memory and delivery yourself, or if the brief should lead to actions inside your accounts.<\/p>\n<h3>Scenario B: a heavier daily research agent<\/h3>\n<p>An account-research agent that runs 20 times every working day. Each run uses a frontier model (claude-sonnet-5-5 through the Agent API) with 30,000 input tokens and 5,000 output tokens, 8 standard web searches and 10 URL fetches.<\/p>\n<table style=\"border-collapse:collapse;width:100%;margin:12px 0 24px\">\n<thead>\n<tr>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Line<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Calculation<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Cost<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Input tokens<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">30,000 \u00d7 $2.00 \/ 1M<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.060<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Output tokens<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">5,000 \u00d7 $10.00 \/ 1M<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.050<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Web searches<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">8 \u00d7 $0.0025<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.020<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">URL fetches<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">10 \u00d7 $0.0005<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$0.005<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\"><strong>Per run<\/strong><\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\"><\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\"><strong>$0.135<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Runs per month<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">20 \u00d7 21.67<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">\u2248 433 runs<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\"><strong>Perplexity API spend<\/strong><\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">433 \u00d7 $0.135<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\"><strong>\u2248 $58.50 a month<\/strong><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><strong>Result:<\/strong> $58.50 sits between Katteb Lite ($19) and Katteb Goals ($99). Whether Lite would cover this workload depends on how many minutes of agent time each run takes, which Katteb shows on the run card before you start; we have not measured that here, so we make no claim that one is cheaper. The useful number is the break-even: at $0.135 a run, $19 buys about 140 API runs a month and $99 buys about 733. Switch the model to perplexity\/sonar ($0.25 \/ $2.50 per million tokens) and the same run costs about $0.045, or roughly $19.50 for 433 runs, almost exactly the price of Lite.<\/p>\n<h3>Scenario C: a subscription for one researcher<\/h3>\n<p>If a single person will do the research in an app, compare subscriptions rather than API calls. Using annual billing for both:<\/p>\n<table style=\"border-collapse:collapse;width:100%;margin:12px 0 24px\">\n<thead>\n<tr>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Plan<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Listed price<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Per year<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Perplexity Pro<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$17\/month, billed annually<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$204<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Katteb Lite<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$190\/year<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$190<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Perplexity Max<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$167\/month, billed annually<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$2,004<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Katteb Goals<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$990\/year<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">$990<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><strong>Result:<\/strong> on annual list prices, Katteb Lite is $14 a year less than Perplexity Pro, and Katteb Goals is $1,014 a year less than Perplexity Max. These are not like-for-like products, though. Perplexity&#8217;s plans are built around a person asking questions, with access to many models, deep research and Perplexity Computer; Katteb&#8217;s plans are built around agents that run tasks and goals on a schedule. Pick by the job, not by the $14.<\/p>\n<div style=\"border:2px solid #7b61ff;border-radius:14px;padding:22px 24px;margin:28px 0;background:#f6f4ff\">\n<p style=\"margin:0 0 8px;font-size:20px\"><strong>Turn your research questions into a schedule<\/strong><\/p>\n<p style=\"margin:0 0 14px\">Describe the research you want done in one sentence, for example &#8220;every Monday, check what our 10 competitors changed on their pricing pages and send me a short brief&#8221;. Katteb writes the plan, then an agent runs it and reports every day.<\/p>\n<p style=\"margin:0\"><a href=\"https:\/\/katteb.com\/chat\/?utm_source=katteb_blog&#038;utm_medium=organic_article&#038;utm_campaign=platform_ai_agents&#038;utm_content=perplexity-ai-agent-pricing\" style=\"display:inline-block;background:#7b61ff;color:#fff;padding:12px 22px;border-radius:10px;text-decoration:none;font-weight:700\">Start my research agent<\/a><\/p>\n<\/div>\n<h2>Which route fits which job?<\/h2>\n<table style=\"border-collapse:collapse;width:100%;margin:12px 0 24px\">\n<thead>\n<tr>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">If you need\u2026<\/th>\n<th style=\"text-align:left;border-bottom:2px solid #7b61ff;padding:8px\">Better fit<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Search-grounded answers inside your own product, at volume<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Perplexity API<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">The lowest cost per question, and you have engineers<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Perplexity API<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">One person doing better research by hand<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Perplexity Pro or Max<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Recurring research that reports changes, with no code<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Agent platform such as Katteb<\/td>\n<\/tr>\n<tr>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Research that leads to actions in signed-in accounts (sheets, CRM, social, email)<\/td>\n<td style=\"padding:8px;border-bottom:1px solid #e5e5f0\">Agent platform such as Katteb<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h2>Five questions to answer before you choose<\/h2>\n<ol>\n<li><strong>How many research runs a month, really?<\/strong> Count them. Below a few hundred light runs, API spend is often a few dollars; the real cost is the code around it.<\/li>\n<li><strong>How deep is each run?<\/strong> One search and a short answer is cheap. Ten searches, ten fetches and a frontier model is roughly 25 times more per run in our examples.<\/li>\n<li><strong>Does the result need to go somewhere?<\/strong> Every destination (email, Slack, a sheet) is an integration you either build or get included.<\/li>\n<li><strong>Does anything happen after the answer?<\/strong> If yes, and it happens inside a site you log in to, you need an agent that can act, not just an API that answers.<\/li>\n<li><strong>Who checks the citations?<\/strong> Search-grounded answers can still be wrong. Name the person who reviews anything that goes to a client or into a decision.<\/li>\n<\/ol>\n<h2>Setting up a research agent as a goal<\/h2>\n<p>If the agent route fits, most of the work is writing a clear goal. A goal is an outcome with a deadline; the agent plans the steps, runs them on schedule, remembers what it found and reports.<\/p>\n<ol>\n<li><strong>Say what should be true.<\/strong> &#8220;By the end of each Monday, I have a one-page brief on what our 10 named competitors changed in pricing, features and messaging, with a link for every claim.&#8221;<\/li>\n<li><strong>Name the sources.<\/strong> List the competitor sites, review sites or news sources you trust. Fewer, better sources beat &#8220;search everything&#8221;.<\/li>\n<li><strong>Say where the output goes.<\/strong> A report in the thread, an email, or a row in a named Google Sheet.<\/li>\n<li><strong>Set the boundary.<\/strong> State what the agent may do on its own (read, summarise, log) and what needs your approval (anything sent to a customer or published).<\/li>\n<li><strong>Review the first two weeks.<\/strong> Read every brief, correct anything wrong, and tighten the goal. After that, read the summary and only open what is flagged.<\/li>\n<\/ol>\n<h2>When you do not need any of this<\/h2>\n<p>Not every research job needs an agent. If you ask a question once a month, open Perplexity or any search engine and ask it. If the answer is a stable fact, write it down once. If you only need the raw search results inside an existing pipeline, the Search API at $5 per 1,000 requests is simpler than any agent. An agent pays for itself when the same questions repeat, the answers change, and someone currently spends hours a week doing the checking by hand.<\/p>\n<h2>Common mistakes<\/h2>\n<ul>\n<li><strong>Budgeting only the API line.<\/strong> The API invoice is often the smallest part of running a research agent; the engineering and maintenance around it is the larger one.<\/li>\n<li><strong>Using the deepest preset for every question.<\/strong> Most recurring checks are single facts. Save multi-step research for questions that need it.<\/li>\n<li><strong>Not storing last week&#8217;s answer.<\/strong> Without memory, every brief repeats everything and nobody reads it by week three.<\/li>\n<li><strong>Treating a citation as proof.<\/strong> A link is where to check, not a guarantee. Spot-check the sources.<\/li>\n<li><strong>Comparing products that do different jobs.<\/strong> An app subscription, a metered API and an agent platform solve different problems; compare them on the job you have.<\/li>\n<\/ul>\n<h2>FAQ<\/h2>\n<h3>How much does the Perplexity API cost?<\/h3>\n<p>It is usage-based. On 8 October 2026 Perplexity listed web search at $2.50 per 1,000 calls ($1.00 with Fast Search), URL fetches at $0.50 per 1,000, the Search API at $5.00 per 1,000 requests, and its Sonar model at $0.25 per million input tokens and $2.50 per million output tokens. Perplexity&#8217;s own example of a light &#8220;low&#8221; preset run comes to $0.0055.<\/p>\n<h3>Is the Perplexity API included in Perplexity Pro?<\/h3>\n<p>The Pro and Max plans are app subscriptions &#8220;for personal use only&#8221;. The API platform is billed separately on usage; check Perplexity&#8217;s billing documentation for any credits that apply to your account.<\/p>\n<h3>Is Katteb cheaper than the Perplexity API?<\/h3>\n<p>Not for light workloads: in our Scenario A the API spend was about $0.60 a month against Katteb Lite&#8217;s $19. For heavier daily research the API spend can pass $19 (about $58.50 in Scenario B), but whether a Katteb plan covers the same work depends on agent minutes per run, which we did not measure. Katteb&#8217;s price includes scheduling, memory, reports and browser actions that the API leaves to you.<\/p>\n<h3>Can a Katteb agent use Perplexity?<\/h3>\n<p>Katteb agents work in a real browser and can open any site you can, including perplexity.ai, but the agent chooses its own AI models for each step. If you specifically need Perplexity&#8217;s API inside your own product, use the API directly.<\/p>\n<h3>What is the cheapest way to start?<\/h3>\n<p>Write down the five questions you ask most often, run them by hand for a week, and time it. If it takes minutes, keep doing it by hand. If it takes hours and repeats, either build on the API or <a href=\"https:\/\/katteb.com\/chat\/?utm_source=katteb_blog&#038;utm_medium=organic_article&#038;utm_campaign=platform_ai_agents&#038;utm_content=perplexity-ai-agent-pricing\">describe the job to a Katteb agent<\/a> and let it run the schedule.<\/p>\n<h2>The bottom line<\/h2>\n<p>Perplexity&#8217;s API is an inexpensive, well-documented way to get cited, current answers into software, and for light research it is hard to beat on raw cost. What it does not provide is the rest of an agent: the schedule, the memory, the delivery and the actions. If you have engineers and a pipeline, build on it. If you want the research done and the follow-up handled without writing code, a goal-driven agent platform is the other route, at a flat monthly price. Run the numbers above with your own volumes, then <a href=\"https:\/\/katteb.com\/chat\/?utm_source=katteb_blog&#038;utm_medium=organic_article&#038;utm_campaign=platform_ai_agents&#038;utm_content=perplexity-ai-agent-pricing\">try the agent route on one recurring research job<\/a> and compare.<\/p>\n<p style=\"font-size:13px;color:#666\">Prices in this article are list prices in US dollars read from docs.perplexity.ai (pricing and Agent API models), perplexity.ai\/pro and katteb.com\/pricing on 8 October 2026, before tax. They change; check the live pages before you decide.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>What a Perplexity AI agent does, the official Perplexity API prices (read 8 October 2026), three worked cost scenarios with every assumption stated, and how to choose between the API, Perplexity Pro and an agent platform.<\/p>\n","protected":false},"author":1,"featured_media":582,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3],"tags":[],"class_list":["post-581","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-guides"],"_links":{"self":[{"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/posts\/581","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/comments?post=581"}],"version-history":[{"count":1,"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/posts\/581\/revisions"}],"predecessor-version":[{"id":583,"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/posts\/581\/revisions\/583"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/media\/582"}],"wp:attachment":[{"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/media?parent=581"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/categories?post=581"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/katteb.com\/blog\/wp-json\/wp\/v2\/tags?post=581"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}