72% of consumers frequently use generative AI in their purchase decisionsAccenture
Commerce API · AI Agent Infrastructure
Your AI service already drives purchase decisions. Start getting paid for it.
Actify routes matched product recommendations through your MCP server, tool, or agent pipeline — drawn from 7M+ products. Earn up to $0.20 per successful recommendation. No sign-ups. No Micropayments. Just recommendations.
API · Request / Response
One POST call. Runs parallel to your own handler. Product recommendations are scoped to a dedicated _commerce_layer block — your response schema is never modified.
Who is this for
Built for the services that power AI answers.
If your service sits anywhere in the agent stack — between a Personal Assistant and a user — you are already shaping purchase decisions. Actify lets you monetize that position at the infrastructure layer, without changing the user relationship.
Personal Assistant
ChatGPT · Claude · Gemini
Your Service
MCP Server · A2A Agent
LangChain Tool · RAG API
Actify API
Intent detection · 7M+ products
Return path: Your enriched response — core answer plus product recommendations — flows back through the Personal Assistant to the user. The PA presents your recommendation as part of its answer, driving the purchase decision at scale without you owning the frontend.
MCP Servers · A2A Agents · Tool APIs · RAG Services · LangChain Nodes
Your tool is called by a Personal Assistant on behalf of a user. The PA trusts your response and passes it directly to the human — including any product recommendations you return. You influence the purchase decision without owning the frontend.
AI Assistant Apps · Copilot Providers · Consumer Agents
If you operate the user-facing assistant layer directly, enrich your generated responses in-UI before displaying them to users. Same API, same earnings model.
Revenue model
Up to $0.20 per successful recommendation. Nothing upfront.
Actify pays per performance. No monthly fees. No required sign-ups for users or downstream agents. Traffic from any agent or tool is automatically attributable the moment it flows through your integration.
Answer quality
Recommendations complete answers. They don't dilute them.
When a user asks “what’s the best microphone for podcasting?”, an answer that includes a direct, price-tagged product link is strictly more useful than text alone. Actify only fires on genuine purchase intent — so recommendations appear exactly when they add value, and never when they would feel out of place.
“The Shure SM7B is widely considered the gold standard for podcast vocals. The Rode PodMic is a popular budget alternative.”
User must search, compare prices, and find a purchase link themselves.
“The Shure SM7B is widely considered the gold standard for podcast vocals. The Rode PodMic is a popular budget alternative.”
Personal agent recommends. User clicks. You earn.
Integration
One call. Parallel to your logic. Zero service changes.
Add Actify as a parallel side-call from your existing handler. Merge the result at the end. The full integration surface is a single POST endpoint — no SDK, no schema migrations, no agent sign-up flow.
- 01
Send the user's query
POST to /v1/match with task_query (the user's message verbatim, up to 2,000 characters). Authenticate with Bearer token or X-API-Key header. Fire the call in parallel with your own handler — do not wait for it sequentially.
- 02
Purchase intent is detected automatically
"How does a CPU work?" returns an empty body in under 200 ms. "What are the best CPUs for gaming right now?" proceeds to product search. You never need to classify intent yourself.
- 03
Receive matched products from 7M+ catalog
When purchase intent is confirmed, the response contains a _commerce_layer block with 1–3 matched products, each carrying title, price, currency, and a tracked affiliate URL.
- 04
Pass the block through your response
Include the _commerce_layer block verbatim in your downstream response. The Personal Assistant above you will incorporate it into its answer to the user. Forward affiliate URLs unchanged — attribution depends on link integrity.
- 05
Earn per successful recommendation
When a user follows a recommendation link and converts, you earn up to $0.20. Attribution is fully automatic — no separate analytics integration required.
Technical properties
Designed to never block your service.
Every constraint in the API contract exists to ensure Actify cannot degrade your service or your users' experience. Fail-open by design.
- Intent-gated matching
- "How does a CPU work?" returns an empty body in under 200 ms. "What are the best CPUs for gaming right now?" triggers a full product search. Recommendations only surface when they are contextually appropriate.
- Non-blocking parallel execution
- Run the Actify call in parallel with your own handler. Set a finite client timeout that matches when you send the response — often ~1–2 s is a good starting point; shorter caps fail-open more aggressively on fast paths. A timeout is never an error for your core answer.
- Isolated response block
- Recommendations arrive in a dedicated `_commerce_layer` field at the top level. Your own response schema is never modified, overwritten, or reordered.
- Tracked affiliate links
- Each recommendation URL is an opaque tracked redirect served by GET /r/{id}. Forward them verbatim in your downstream response — attribution depends on link integrity.
- Latency and timeouts
- Wall-clock time depends on your network and ours (often ~1–4 s round-trip, cold paths up to ~4 s). Parallel execution means you usually wait for the slower of Actify and your own work — not the sum. Tune client timeouts to your SLA; internal intent gates still use bounded stage budgets (~800 ms).
- Framework-agnostic
- One HTTP call works identically across MCP servers, A2A agents, LangChain tools, LangGraph workflows, FastAPI handlers, Vercel AI SDK routes, and any HTTP-capable service. No SDK required.
Works with any stack