Cognatu.Cognatu.Cognatu matches contextual ads to your LLM app's responses — on your device, without sharing user data. One line of middleware. Revenue from day one.
Your free users cost you inference money on every message. One contextual ad per conversation turns your biggest cost centre into your first revenue line — no subscription wall needed.
Monetise the free tier of your coding assistant, CLI, or IDE plugin. Ads match what developers are actually working on — frameworks, deploy targets, infra — so they read as recommendations, not noise.
Embedded product assistants answer thousands of questions a month. Relevant sponsor placements pay for the tokens, and your users keep getting free answers.
Add the Cognatu middleware in front of your existing OpenAI, Anthropic, or any LLM call. Works with streaming responses.
A 22 MB MiniLM model extracts topics and segments from the conversation locally. Only an anonymised context vector leaves your server.
Winning ads are returned with the LLM response. Earnings accumulate in your publisher portal — withdraw at any time.
// Wrap your existing chat handler with Cognatu middleware. import cognatu "github.com/shirishgarg/cognatu/pkg/sdk" func main() { client := cognatu.New("YOUR_PUBLISHER_KEY") mux.Handle("/chat", client.Middleware(yourChatHandler)) }
# Drop into any FastAPI / Django / Flask app. from cognatu import CognatuMiddleware app.add_middleware( CognatuMiddleware, publisher_key="YOUR_PUBLISHER_KEY", )
// Add one script tag — works on any page with LLM-generated content. <script src="https://gateway.cognatu.com/sdk/v1.js" data-publisher-key="YOUR_PUBLISHER_KEY" data-slot="#ad-slot" async> </script>
# POST the conversation context — get back a winning ad. curl -X POST https://gateway.cognatu.com/v1/auction \ -H "X-Publisher-Key: YOUR_PUBLISHER_KEY" \ -d '{"messages": [...], "llm_text": "..."}'
Topic extraction runs on your server via an embedded 22 MB MiniLM model. Users' words never leave your infrastructure.
Ads match the actual topic of the AI response — not the page title or a third-party cookie profile. Relevance = higher CPMs.
The full auction waterfall — sponsorships, paced delivery, second-price network — completes in under 45ms so it never slows your LLM call.
See hourly fill rate, lost-auction reasons, and CPM trends in your publisher portal. Know exactly what's earning and what isn't.
Run split tests on ad formats, placements, and creative copy. Built-in variant assignment and per-experiment stats.
Advertisers optionally target users with verifiable audience proofs (Groth16 ZKP). Higher CPM for publishers whose users qualify.