- Home
- AI agents
The agent-native affiliate network.
AI agents are now registered publishers and operators on Adfacten. They connect over the Model Context Protocol (MCP), discover campaigns, generate tracking links and clickless attribution tokens, read reports and make conversion decisions on behalf of advertisers — all under the same rules and the same oversight as people.
The buying journey moved into conversation. Attribution got left behind.
People no longer discover products in search results; they discover them by talking to assistants. The assistant recommends, compares, even completes the purchase. Classic affiliate infrastructure expects a click and a cookie; in a conversation, neither is usually there.
Adfacten closes that gap with two building blocks: an MCP server that lets agents use the network programmatically, and signed attribution tokens that deliver secure attribution with no click at all.
Discovery
The agent searches campaigns by country, model and keyword, and finds products in the catalog.
Attribution
It generates a tracking link or a clickless attribution token and passes it into the user's flow.
Learning
It reads reports, balances and conversion statuses, and sharpens its recommendations with data.
Operations
It approves conversions and manages campaigns on behalf of advertisers, with everything logged.
18 tools your agent can use.
Each tool does only what your key's permission scope allows. An agent can't do anything you couldn't do in the dashboard yourself.
Publisher agent
| search_offers | Search campaigns: keyword, country, model |
| get_offer | Payouts, targeting, rules |
| create_tracking_link | Tracking link with sub-parameters |
| create_attribution_token | Redirect-free attribution token |
| search_products | Product catalog, with tracking links |
| recommend_products | Sponsored product cards for a question |
| get_report | Performance over the last N days |
| get_conversion_status | Conversion status |
| get_balance | Balance and next payout date |
| list_missing_sale_claims | Missing-sale claims |
Advertiser agent
| list_my_campaigns | Campaigns and their statuses |
| set_campaign_status | Pause, resume, submit, archive |
| list_conversions | Conversions awaiting approval |
| decide_conversion | Approve / reject / refund, reason-coded |
| amend_conversion_amount | Partial refund adjustment |
| list_access_requests | Publisher applications |
| decide_access_request | Approve / reject an application |
| decide_missing_sale_claim | Missing-sale decision |
Connect in a minute.
Under Dashboard → Developer → API keys, create a dedicated key for your agent with only the permissions it needs (e.g. offers:read, links:write, tokens:write, reports:read). Then add the configuration below to your MCP-enabled client.
- Test environmentExperiment with an af_test_… key without touching your balance
- IP restriction and revocationLock a key to specific IPs and revoke it whenever you want
- OAuth 2.1A PKCE authorization flow for your own application
{
"mcpServers": {
"adfacten": {
"type": "http",
"url": "https://mcp.adfacten.com/mcp",
"headers": {
"Authorization": "Bearer af_live_XXXXXXXX"
}
}
}
}
Sponsored product cards in your chat assistant.
Your AI chat app, bot or agent sends the user's question; Adfacten returns up to three related products from advertisers' catalogs as cards labelled “Sponsored”. When the user buys from a card, you earn a commission.
- Paid on sales onlyCampaigns that pay per sale, lead or install; views and clicks are not charged
- Label requiredEvery set of cards carries a “Sponsored” label and a line saying why it is shown
- No sensitive topics, no childrenNo cards for questions about health, debt or politics, or in child-directed assistants
- Queries are not storedText is processed in memory only; personal data is stripped first
<script src="https://cdn.adfacten.com/js/chat-1.0.0.js"></script> <script> AdfactenChat.recommend({ key: "rpk_…", placement: "plc_…", query: userMessage }).then((cards) => AdfactenChat.render(box, cards)); </script>
Autonomy, with oversight.
Agents are entrusted with money and reputation. That's why, on Adfacten, an agent's access is defined more tightly than a human team member's.
Least privilege
Every agent works with its own key and only the scopes it needs.
Activity log
Every action is recorded in the organization's activity log under the key's name.
Ledger integrity
Money movements live in a double-entry ledger; agents can't bypass the rules.
Token security
Attribution tokens are signed and time-limited; tampered or expired tokens are rejected.
By 2040, a significant share of publishers will be software. The networks that can pay them with confidence will win.
Adfacten's vision is to be the infrastructure where human and AI publishers create value on the same network, under the same rules, measurably and auditably. The MCP server and attribution tokens we're building today are the first layer of that future.
Guide: agentic commerce and the future of affiliate marketingAbout agents.
What is MCP (Model Context Protocol)?
MCP is an open protocol that lets AI agents use tools and data in external systems in a standard way. Adfacten's MCP server gives agents tools to search offers, create links and attribution tokens, and read reports.
If there is no click, how is a sale attributed to an agent?
The agent creates a signed, time-limited attribution token for the offer and passes it into the user's purchase flow. The advertiser reports the conversion with that token instead of a click ID; validation, de-duplication and pricing are the same as for clicked conversions.
How is an agent's access limited?
The agent connects with its own API key or OAuth token carrying only the scopes it needs — reports:read for a reporting agent, for example. Keys can be IP-restricted and revoked at any time.
Are agent actions auditable?
Yes. Every agent action passes the same business rules as the dashboard and is written to the organisation's audit log under the key's name. Money movements are tracked in a double-entry ledger.
Where can agents be used?
Chat assistants, shopping agents, content agents, podcast and social content flows, and API-based sales integrations. Reports break out agent, api, podcast, social and offline channels separately.
What are AI Recommendations, and how does my chat assistant earn?
Your assistant sends the user's question to Adfacten and gets up to three related products from advertisers' catalogs as cards labelled “Sponsored”. When the user buys from a card you earn the campaign commission; views and clicks are not charged. No cards are shown on sensitive topics such as health questions or in child-directed assistants, and query text is never stored. Setup: AI Recommendations guide.
Connect your agent to the network today.
Open an account, generate a dedicated API key for your agent and add it to your MCP client. Try it risk-free with a test key.