Every free user costs you tokens. Adgentek is the ad server for AI apps, chatbots, and agents: native ads that render in your own UI, clearly labeled, matched to what the conversation is about. Buyers get derived signals only, never raw conversations.
Nothing is injected into your model's output. The unit renders in a slot you place, in your DOM, with your styling, and carries a visible Sponsored label. Users always know what is an ad and what is the answer.
Adgentek uses the conversation to make the decision. Advertisers receive an intent category, a decision stage, a topic, and aggregate performance. No prompts, no transcripts, no PII, no user profiles, no cookies.
docs.adgentek.ai publishes an integration skill and an llms.txt index. Paste one prompt into Claude Code, Cursor, or Codex. The agent drops in the JavaScript SDK, places the slot, and asks you for your key. Most apps are live the same day.
You built an AI app people use. Every turn runs a model call, and that call has a price whether the user is on a paid plan or not. Most are not. The usual answers are a paywall, a message cap, or a cheaper model for free users, and each one costs you the growth that made the app worth building.
Meanwhile a meaningful share of those same conversations carry real commercial intent. Someone comparing options, pricing something out, planning a purchase, choosing a vendor. In a browser those are the most valuable moments in advertising. Inside your app they currently earn nothing.
That is what Adgentek is. An ad server built for AI apps, chatbots, and agents, connected to four tiers of demand: direct campaigns sold against your inventory, programmatic demand from leading partners, CPC feeds, and outcome-priced budgets. A request that does not clear the top tier keeps falling until it finds a buyer, rather than coming back empty.
The format is native to a conversation rather than borrowed from a web page, because a banner forced into a message thread damages the trust you spent months earning. For the full breakdown of the three models, see ad networks vs ad servers for AI apps.
The ad request is dispatched in parallel with your completion, not ahead of it, so your response streams on its own schedule and the slot fills when the decision returns. If it does not return in time, the slot stays empty and the conversation is unchanged. An ad server outage is invisible to your users by design, and you can verify that by pointing the SDK at a dead endpoint. Covering the inference bill is not worth a slower app, so the ad path was built to never be in the way.
Generate a slot ID, then paste the install prompt below into your agent. It reads docs.adgentek.ai, drops in the JavaScript SDK, and places the slot. Prefer to do it by hand? The SDK is a few lines. Get Started Free →
You decide where the unit appears: after the first response, every Nth turn, only when the conversation turns commercial. It renders in your DOM with your styling. Nothing touches the model's answer.
Each request is classified in real time into a topic, an intent category, and a decision stage. The raw text is used for the decision and is not forwarded to buyers.
Direct, programmatic, CPC, and outcome-priced buyers bid against the derived signal. A request that does not clear the top tier keeps falling until it finds a buyer.
A Sponsored unit appears in your slot: a contextual card, an interactive Q&A, a sponsored recommendation. Users see what is an ad and what is the answer, and they engage because it matches what they were already asking about.
Reporting shows fill, eCPM, and revenue by slot. Set a quality floor, raise it as fill grows, and watch the line move against your model bill.
If you use a coding agent, start with the install prompt. Below it, the same integration by hand.
Integrate Adgentek ads into this app using the JavaScript Web SDK. Start by fetching https://docs.adgentek.ai/skill.md and following it, and fetch https://docs.adgentek.ai/llms.txt to discover the current documentation pages. Load the SDK client-side, call AdgentekAds.requestAd after each assistant response, and render the ad in a slot in our own UI with the Sponsored label intact. Do not invent SDK methods; verify every symbol against the live docs. Ask me for the real publisher API key and slot ID rather than leaving placeholders, and never fire an impression pixel from server-side code.
Works with Claude Code, Cursor, Codex, and any agent that can fetch a URL. Complete documentation is at docs.adgentek.ai. Agents should start from the Adgentek integration skill and use the docs index rather than a cached snippet.
<script src="https://api.adgentek.ai/storage/v1/object/public/sdk/sdk.js"></script>
<script>
AdgentekAds.init({ apiKey: 'YOUR_FULL_API_KEY_HERE', slotId: 'YOUR_SLOT_ID' });
</script>
// the slot lives in your UI, styled by you, labeled Sponsored
<div id="adgentek-ad"></div>
// call after each AI response
AdgentekAds.requestAd([
{ role: 'user', content: userMessage },
{ role: 'assistant', content: aiResponse }
]);
Your API key is generated in the dashboard under Settings, API Keys and is prefixed adgt_. The slot ID is the placement ID from Inventory, Surfaces, and it drives targeting, floor pricing, and reporting. Placeholder credentials do not serve ads. React, iOS, and Android SDKs are in development.
The SDK is for apps where the user sees a screen: web chat, assistants, vertical AI tools. If your product runs headless, as an agent, a voice surface, or a backend service, talk to us and we will set you up directly. Contact Us
Each unit renders in your UI with a visible Sponsored label and is designed to read as a useful next step, not an interruption. Users experience them as helpful because they match what the user was already asking about.
| Format | Description | Best For | Engagement |
|---|---|---|---|
| Contextual Cards | Native cards that surface relevant products or services based on conversation topic | Product discovery, recommendations | High visibility, low friction |
| Interactive Q&A | Brand-authored guided conversation with chip selections and free-text, users self-qualify through Q&A | Considered purchases, lead gen | 3.2 avg interactions, deep intent |
| Sponsored Recommendations | Contextually-matched product suggestions that appear as natural conversation extensions | E-commerce, travel, services | Natural discovery, high relevance |
| Inline Mentions | Brief, relevant sponsored references woven contextually into AI responses | Brand awareness, consideration | Non-disruptive, high reach |
| Action Cards | Conversion-focused units with direct CTA, book, buy, schedule, locate | High-intent moments, bottom funnel | Direct conversion path |
Adgentek's founders have been building programmatic infrastructure since 2007, one on the sell side and one on the buy side.
Founded Beachfront in 2007 and built it into a cross-screen video SSP serving CTV and streaming publishers. Acquired by Seedtag in 2024. Seventeen years on the sell side.
Co-founder and Chief Technology Officer of Simpli.fi, where he built the programmatic DSP and the buy-side infrastructure underneath it. Two decades on the demand side.
Yield management, waterfall logic, floor pricing, discrepancy handling, and the demand relationships that make a fill rate real are not new problems. They are the problems this team spent two decades solving before AI surfaces existed. Adgentek is that stack rebuilt for a conversational surface, not a first attempt at ad serving.
A single direct-sold demand pool leaves most requests unfilled. Adgentek runs a hard waterfall, so a request that does not clear the direct tier keeps falling until it finds a buyer rather than returning empty.
Your assumptions, our math. Adjust every field to match your traffic and your model spend.
Illustrative only. Actual eCPM and fill depend on your topic mix, geography, render path, and format selection. Conversations with commercial intent carry most of the revenue, so apps in travel, shopping, finance, home, and auto tend to cover more of the bill.
Adgentek reads the conversation to make the decision. Buyers get what they need to bid and nothing they could read. No cookies, no user profiles, no PII, no transcripts.
Real-time analysis of conversation meaning, not keyword matching. "Family SUV under $50K" becomes an intent signal for buyers. The sentence itself never leaves the decision.
Knows when a user is browsing vs. comparing vs. ready to buy. Serves the right ad format at the right moment in the decision journey.
Maps conversation topics to advertiser categories. Auto discussion triggers auto ads. Mattress question triggers bedding ads. Precision without profiles.
Automatically excludes sensitive topics, health, mental health, politics, financial distress. Ads never appear in inappropriate contexts.
We're infrastructure. You decide how monetization works in your product.
Choose which ad formats appear in your app. Cards only? Q&A only? Mix and match per placement.
Define exactly where ads appear in the conversation flow. After first response? Every 5th message? On specific intents?
Allow or block entire advertiser categories. No gambling in your education app. No alcohol in your health bot.
Block specific brands or competitors from appearing in your app. Period.
Set maximum ad exposure per session, per user, per day. Protect the experience at your threshold.
Set minimum CPM or engagement thresholds. Only premium advertisers that meet your standards.
Conversational search apps where users ask questions and get synthesized answers. Contextual ads match the query and appear alongside the response.
General-purpose AI chat apps. Ads surface when the conversation touches commercial topics, naturally, contextually, non-intrusively.
Category-specific AI tools, travel planners, recipe assistants, fitness coaches, financial advisors. Deep category relevance drives high ad performance.
Entertainment and social AI experiences. Sponsored recommendations and contextual cards that fit the conversation genre and audience.
Each surface has a different latency budget, session shape, and set of formats that fit. These go deeper than this page does.
Consumer and vertical chat apps. Full format set, certified client render, programmatic CPM eligible.
Read the guideIDE assistants, terminal agents, autonomous runs. Category allowlists and a strict non-blocking request path.
Read the guideSpark for Publishers answers reader questions from your own archive at the end of every article. Hold the reader, deepen the session, and monetize the follow-up question.
Read the guideNot when they are native, labeled, and relevant. The unit matches what the user is already asking about, carries a visible Sponsored label, and renders in your UI on your terms: which formats, where in the flow, how often. You set the frequency cap and the quality floor. Users leave apps that paywall them or cap their messages. A labeled card after a commercial question is a far smaller ask.
No. Nothing is injected into the model output. The ad renders in a slot you place in your own interface, with your styling, and it is clearly labeled Sponsored. That keeps the answer trustworthy, and rendering client-side is also what qualifies your inventory for impression-priced direct and programmatic demand.
Derived signals only: an intent category, a decision stage, a topic, and aggregate performance metrics such as impressions, engagement, and conversions. Never raw conversations, chat history, prompts, personal details, or user profiles. Buyers bid on the signal, not the transcript.
Yes. docs.adgentek.ai publishes an integration skill at skill.md and a docs index at llms.txt. Create a free account, generate a slot ID, and paste the install prompt above into Claude Code, Cursor, Codex, or any agent that can fetch a URL. It drops in the JavaScript SDK, places the slot, and asks you for the real key. Get Started Free →
It depends on volume, topic mix, and render path. Use the estimator above with your own numbers. Conversations with commercial intent carry most of the revenue, so apps where users compare, price, and choose cover more of the bill than general chat. Our interactive formats generate higher engagement than static display, which is what moves effective CPMs. We share specific projections during onboarding based on your traffic profile.
No. The ad request runs in parallel with your completion, never in front of it. If the decision does not return in time, the slot stays empty and the conversation is unchanged.
Full control. Block categories, block specific brands, set quality floors, approve or reject individual advertisers. No gambling in your education app. No alcohol in your health bot. You decide what appears in your product.
Yes, but not through the SDK on this page. The Web SDK is for apps where the user sees a screen. If your product runs headless, as an agent, a voice surface, or a backend service, Contact Us and we will set you up directly. See also what conversational advertising is.
Buyers receive intent, stage, topic, and aggregate performance. Never raw conversations, PII, or user profiles.
Sensitive topic exclusion. Category controls. Brand blocklists. Advertiser quality review.
Every unit renders in your UI with a visible Sponsored label. Users always know what is an ad and what is the answer.
GDPR, CCPA, and emerging AI advertising regulations. Designed for the regulatory environment ahead.
Create a free account, generate a slot ID, and paste one snippet or hand the install prompt to your coding agent. Labeled native ads in your own UI, live today. No sales call.
Founding Member, Ad Context ProtocolBuilt by the founders of Beachfront and Simpli.fiBy serving labeled native ads inside the app. Adgentek is an ad server built for AI apps, chatbots, and agents. It classifies each conversation in real time, runs four tiers of demand (direct, programmatic, CPC, and outcome-priced) against that signal, and renders a clearly labeled native unit in the publisher's own UI. Revenue from commercial-intent conversations offsets the model cost of serving free users, so the app does not need a paywall or a message cap.
No. Adgentek never injects text into the model output. The ad renders in a slot the publisher places in their own interface, with their own styling, and it carries a visible Sponsored label. Users can always tell what is an ad and what is the answer. Rendering client-side in the publisher's UI is also what qualifies the inventory for impression-priced direct and programmatic demand.
Derived signals only. Advertisers receive an intent category, a decision stage, a topic, and aggregate performance metrics such as impressions, engagement, and conversions. They never receive raw conversations, chat history, prompts, personal details, or user profiles. Adgentek uses the conversation to make the ad decision; buyers bid on the signal, not the transcript.
Yes. docs.adgentek.ai publishes an integration skill at docs.adgentek.ai/skill.md and a documentation index at docs.adgentek.ai/llms.txt. A publisher creates a free account, generates a slot ID, and pastes the install prompt from this page into Claude Code, Cursor, Codex, or any coding agent. The agent reads the docs, drops the JavaScript SDK into the app, places the slot, and asks for the real API key. Most apps are live the same day.
Full control. Publishers choose which ad formats appear, where they are placed in the conversation, which advertiser categories are allowed or blocked, minimum quality thresholds, frequency limits, and brand restrictions. Adgentek operates as infrastructure and the publisher decides the experience. Monetization can be paused for specific user segments, and all active advertisers can be reviewed.