Full Ad Server for AI Applications

Ad revenue for AI apps — one tool call

AdsMCP is a full ad server hosted as an MCP. Serve live programmatic demand, bring your own direct-sold campaigns, or both. Your AI app calls ads.serve. Adgentek handles everything else.

Create Publisher Account View on GitHub

mcp-config.json

        // 1. Add to your MCP client config
        {
        "mcpServers": {
        "adgentek": {
        "command": "npx",
        "args": ["@adgentek/adsmcp-server"],
        "env": {
        "ADGENTEK_API_KEY": "YOUR_API_KEY"
        }
        }
        }
        }
        

        // 2. Call ads.serve in your app
        const result = await callTool('ads.serve', {
        messages: conversationHistory,
        surface: 'conversational'
        });
        

        // 3. Tracking is fully automatic
      

Create your publisher account

Add the MCP server config

One JSON block in your MCP client. No SDK, no package install for the remote option.

Call ads.serve and earn

Pass conversation context. Render the response. All tracking is automatic — no event calls required.

MCP Tools

Four tools. Complete ad server control.

AdsMCP exposes exactly the tools you need — from serving and prefetching to configuration and diagnostics.

ads.serve Primary

Serves a contextual ad based on conversation content. AdsMCP analyzes messages for topic, intent, sentiment, and depth — then matches and returns the highest-yield ad. Impressions, clicks, and conversions are tracked automatically. No manual event calls.

messages surface sessionId userKey intent placementCandidates

ads.prefetch Performance

Batch-request multiple ads for caching and reduced latency at placement time. Pass an array of context objects, get back an array of ready-to-render ads. Cache results for up to 5 minutes.

batch[] messages surface sessionId

ads.config Configuration

Returns your publisher configuration, slot definitions, and ad policies. Use this to check slot capabilities (voice, sidebar, native), adapt your integration, and enforce brand safety policies.

includeSlots includePolicies includeAnalytics

ads.diagnostics Debug

Explains exactly why ads are or are not serving. Returns eligibility checks, per-ad filter reasons (geo mismatch, frequency cap, budget exhausted), and a full auction log including winner, total bids, and floor price.

checkEligibility includeAuctionLog messages surface

Full Ad Server

Bring your own ads — or use ours

AdsMCP is not a demand source with an MCP interface bolted on. It is a complete ad server — purpose-built for conversational AI surfaces, and flexible enough to serve as the single monetization layer for your entire app.

Use Adgentek's demand relationships, bring your own direct-sold campaigns, or connect additional ad sources alongside ours. Yield is optimized automatically across everything — no manual prioritization required.

Adgentek's live demand — established relationships across programmatic, performance, and direct channels, with more added continuously

Your direct-sold campaigns — upload advertiser deals and serve them through the same ads.serve call

Additional ad sources — connect your own demand partners alongside Adgentek's and let the server optimize yield across all of them

What you get

Ad server, not just a network

Full ad serving infrastructure — trafficking, targeting, capping, pacing, and reporting — hosted and managed by Adgentek.

Demand from day one

Start earning immediately. No demand negotiations required — Adgentek's existing relationships fill inventory from your first impression.

Open to your own sources

Already have an advertiser relationship or a preferred demand partner? Bring them in. AdsMCP serves as the unified layer across all sources.

Why AdsMCP

Purpose-built for conversational AI

Traditional ad stacks were built for pages and feeds. AdsMCP is built for the conversational context — where intent is explicit and ad relevance is structurally higher.

MCP-native architecture

A proper remote MCP server — not a REST API with a wrapper. Your agent communicates via Model Context Protocol natively, with full tool-use semantics and streaming support.

Automatic tracking

Impressions, clicks, and conversions are tracked by Adgentek — not the publisher. This eliminates implementation complexity and prevents fraud by design.

Open source reference

Schema, documentation, and integration examples are open at github.com/Adgentek/adsmcp. Proprietary server code runs on Cloudflare Workers.

Intent-aware matching

Nine-bucket intent classification maps any conversation to a standardized signal automatically — purchase, transactional, comparison, research, informational, and more.

Multi-surface support

Built for conversational, search, voice, creative, and web surfaces. Slot capabilities (sidebar, voice, native) are configurable per placement via ads.config.

Revenue share model

Publishers receive a share of net ad revenue on every impression. No upfront costs, no minimum traffic requirements. Earnings tracked to your publisher ID.

User conversations stay private. Full stop.

Raw conversation content never leaves your application and is never shared with advertisers, demand partners, or third parties. Ad matching runs on derived signals only — the minimum information required to serve a relevant ad.

What advertisers receive	What stays in your app
IAB content category	Full conversation content
Intent classification	User messages
Keyword themes	Session context

Conversations never shared beyond Adgentek. User conversation content is not stored, forwarded to demand partners, or used for cross-publisher profiling.

Compliant by design. AdsMCP's signal architecture aligns with emerging AI platform data policies and consumer privacy expectations — built in from the foundation, not added as a layer.

Derived signals only, always. Whether matching on contextual or audience signals, only derived classifications reach demand partners — never raw conversation data.