Ad Buyer Agent¶
The Ad Buyer Agent is an automated advertising buying system built on CrewAI and the IAB OpenDirect 2.1 protocol. It receives a campaign brief, allocates budget across channels, researches seller inventory, builds recommendations, and books deals -- all through a single API.
Part of the IAB Tech Lab Agent Ecosystem -- see also the Seller Agent.
Key Capabilities¶
- Campaign briefing -- accept structured campaign briefs with objectives, budget, dates, audience, and KPIs.
- Budget allocation -- a portfolio-manager agent splits budget across channels (branding, CTV, mobile, performance).
- Inventory research -- channel-specialist agents query seller product catalogs via OpenDirect.
- Recommendation consolidation -- recommendations from all channels are ranked and presented for review.
- Human approval -- optional approval checkpoint before committing spend.
- Deal booking -- approved recommendations are booked as line items against seller APIs.
Communication Protocols¶
The buyer agent communicates with seller agents using three protocols:
| Protocol | Use Case | Speed |
|---|---|---|
| MCP | Automated tool calls -- structured, deterministic | Fast |
| A2A | Conversational discovery & negotiation | Moderate |
| REST | Operator dashboards, legacy integration | Fast |
CrewAI tools use MCP by default. A2A is used for discovery and complex negotiations. See Protocol Overview for detailed comparison.
API Endpoints¶
The buyer agent exposes 7 endpoints across 3 categories:
| Category | Endpoints |
|---|---|
| Health | GET /health |
| Bookings | POST /bookings, GET /bookings/{job_id}, POST /bookings/{job_id}/approve, POST /bookings/{job_id}/approve-all, GET /bookings |
| Products | POST /products/search |
See the API Overview for full details.
Documentation¶
- Quickstart -- install, configure, and run your first booking
- API Reference -- all endpoints, models, and curl examples
- Architecture -- system design, agent hierarchy, and flow diagrams
- MCP Client -- structured tool calls to seller agents
- A2A Client -- conversational discovery and negotiation
- Protocol Overview -- comparison of MCP, A2A, and REST
- Integration -- connecting to seller agents and the OpenDirect protocol