# Introducing Shaped for Agents — retrieval that gets it right the first time

## Semantic search with AI‑native behavioral reranking
Harness behavior to rank search results at a user level, increasing engagement conversion and retention.

### Integrate Easily
Integrate with multiple data sources or type  
[See Code](https://docs.shaped.ai/docs/overview/how-it-works)

### Iterate Rapidly
Easy to use experimentation tooling  
[See Code](https://docs.shaped.ai/docs/model_management/evaluating-your-model)

### SOTA models
Use our model library or bring your own  
[See Code](https://docs.shaped.ai/docs/model_library/overview)

### Self-Serve Fine Tuning
Easily Fine-Tune your own models  
[See Code](https://docs.shaped.ai/docs/feature_types/language#step-3-fine-tuning-models-for-your-task-and-data)

### Modular Infrastructure
Choose your re-ranker, orderer and scorer  
[See Code](https://docs.shaped.ai/docs/overview/architecture#tag/Model-Inference/operation/post_create_user_embedding_models__model_name__create_user_embedding_post)

### Scale Seamlessly
Automatically scale with your request volume  
[See Code](https://docs.shaped.ai/docs/model_management/handling-scale)

## Why unify search and recommendations?

### No extra cost
Fees optimized for unification  
Integrating Shaped for both search and recommendations comes at no extra cost. Shaped’s pricing structure is modular, with a fixed platform fee and tiered usage fees.

### Simplified Architecture
Less complexity, less headaches  
Unifying search and recommendation infrastructure means you get world-class stability, reliability, scalability, and latency that is easier to integrate and maintain.

### Continuous Improvement Transfer
Systems built to enhance each other  
Knowledge and data sharing between unified systems results in continuous improvement and transfer learning, where improved systems generate richer data, and richer data improves systems.

# Get up and running with one engineer in one sprint
Guaranteed lift within your first 30 days or your money back
