Unified query layer
Compose retrieval, ranking, filtering, scoring, and reordering in a single query through ShapedQL, rather than stitching together separate services.
Shaped
Freemium
Real-time context engine for agentic AI, search, and recommendations
What is Shaped?
Shaped is a real-time context engine for agentic AI, search, and recommendations. It combines retrieval, ranking, embeddings, and learning so applications can return relevant results based on user, item, and query context instead of relying on static vector search alone.
The site positions Shaped as a managed alternative to building a retrieval stack from separate databases, ranking systems, connectors, and pipelines. It is used for personalized search, product recommendations, agent memory and retrieval, and other ranking-driven experiences where relevance needs to improve from interaction data.
What can Shaped do?
Hybrid search and ranking
Blend semantic search, keyword search, and behavioral ranking so results can match intent while still honoring exact matches and business rules.
Context-aware personalization
Use user, item, or text context to personalize outputs, including recommendations and agent retrieval that adapt to who is asking.
Native data connectors
Connect batch and streaming sources through native connectors so data freshness and event signals can flow into the model.
Learning from interactions
Train and update ranking models from interaction signals, with the site describing continuous learning and a feedback loop.
Multiple integration paths
Access the product through Python SDK, TypeScript SDK, or MCP, alongside the query interface shown on the site.
Use Cases
“Personalized product recommendations”
For e-commerce teams, Shaped can personalize home pages, product detail pages, carts, email, and post-purchase surfaces using catalog metadata and user behavior.
“Hybrid search”
For product search, it combines semantic and keyword signals so teams can improve relevance, reduce zero-result queries, and still respect exact-match behavior.
“Agent retrieval and memory”
For agentic applications, it can retrieve context by text, user ID, or item ID so assistants can ground responses in the right data.
“Personalized ranking across surfaces”
For content and commerce experiences, it supports feeds, similar-items ranking, category ranking, and item reranking across surfaces.
“Rapid experimentation”
For teams that need to iterate quickly, it provides a managed workflow for connecting data, tuning relevance, and shipping experiments without assembling infrastructure first.
Frequently Asked Questions
How do you get started with Shaped?
Shaped is set up by connecting your data, then configuring a feed or query in its interface. The source materials describe a three-step workflow: connect data, configure the feed or model, and deploy automatically.
What kinds of workflows does Shaped support?
The product supports retrieval and ranking workflows for search, recommendations, agent retrieval, and other personalized surfaces. The source also shows it can be used through a SQL-like query layer and via Python SDK, TypeScript SDK, or MCP.
Does Shaped offer a free tier or trial?
Pricing pages show a free start option with $100 included and no credit card required, plus paid Starter, Standard, and Enterprise plans. Standard is described as usage-based and Enterprise adds custom contracts and dedicated support.
What support options are available?
The public materials mention support during business hours for Standard, and 24/7 support plus a private Slack or email channel for Enterprise. Starter includes Docs + Slack support.
Quick Facts
- Category
- AI infrastructure
- Primary use cases
- Search, recommendations, and agent retrieval
- Interface
- ShapedQL query layer plus Python SDK, TypeScript SDK, and MCP
- Data ingestion
- Native connectors with batch and streaming support
- Pricing
- Free start available; paid Starter, Standard, and Enterprise plans
- Source domain
- shaped.ai
Shaped Traffic Analysis
Traffic data is for reference only.
- Domain Rating
- 64
