Shaped
Enhance ranking systems with real-time AI-powered tools.. [Paid]
Last verified:
What is Shaped?
Shaped is an end-to-end relevance engine designed for real-time personalization and agent memory. It is a unified, queryable relevance engine that connects your data, trains models, and retrieves relevant results in milliseconds by querying text, user ID, or session context. Unlike traditional vector databases, Shaped includes a built-in feedback loop where every interaction makes results better, and it knows who is asking to personalize results per user.
Key features include ShapedQL (a SQL interface for real-time retrieval and ranking), hybrid search combining keyword and semantic relevance, multi-stage ranking pipelines with filtering and reordering, dynamic embeddings with continuous learning, and 30+ native connectors for data warehouses, analytics tools, and streaming platforms. The platform supports use cases like personalized content feeds, search and discovery, AI agent retrieval, similar item recommendations, personalized email, and AI assistants. It delivers <50ms query latency and handles 5B+ documents with 99.95% uptime.
Shaped is built for product and engineering teams who need to drive engagement, conversion, and revenue through personalization. It is ideal for companies building recommendation systems, search features, AI agents with contextual memory, or any application requiring real-time relevance. Teams can connect data, build their first model, and deploy to production in under 7 days without infrastructure overhead.
Shaped pricing
Pricing model: Freemium
$100 free credits with no credit card required at signup. Pricing starts at $0.03 per answer for agent usage (50x cheaper than traditional stacks at $1.50 per answer). The website mentions flat-fee monthly pricing based on number of users and forecasted usage with self-serve and white-glove support tiers. No credit card required for free tier. Contact sales for enterprise pricing, custom deployment needs, and detailed plan limits.
Shaped pros
- <50ms query latency for real-time retrieval
- Built-in feedback loop that improves with every interaction
- $100 free credits with no credit card required
- 30+ native connectors for warehouses, analytics, and streaming
- Unified context in one call instead of 5+ services
- Personalized results per user, not just document similarity
- Hybrid search blending semantic and keyword relevance
- Continuous learning with dynamic embeddings
- 50x cheaper than traditional agent stacks ($0.03 vs $1.50 per answer)
- No retry loops - right results first time
- SOC 2 Type II certified, GDPR/HIPAA compliant
- 99.95% uptime SLA with enterprise-scale reliability
- Python SDK, TypeScript SDK, and MCP support
- Handles 5B+ documents with 1,000+ QPS
- Time to first experiment in ~7 days
- Multi-stage ranking pipeline with filtering and reordering
- No infrastructure required - fully managed
Shaped cons
- Pricing details not publicly listed - requires contact for enterprise
- Newer platform with smaller community than established vector DBs
- Primarily focused on personalization use cases, not general vector search
- Requires data connection setup before getting started
- Self-serve tier may have limited features vs white-glove support
- No mention of on-premise deployment option
- Learning curve for ShapedQL SQL interface
- Heavy emphasis on behavioral signals may not suit all use cases
Frequently asked questions about Shaped
What is Shaped?
Shaped is an end-to-end relevance engine designed for real-time personalization and agent memory. It is the only vector database with a feedback loop, where results are personalized per user and get better with every interaction. Shaped unifies retrieval, ranking, and learning in one query.
How is Shaped different from a vector database?
Unlike a vector store, Shaped knows who's asking and personalizes results per user. Traditional vector databases use static embeddings that never learn, while Shaped has continuous learning with dynamic embeddings. Shaped also unifies retrieval, ranking, and learning in one query instead of requiring 5 separate services.
What use cases does Shaped support?
Shaped powers recommendations, search, and agent retrieval across your entire application. Specific use cases include for-you feeds, search and discovery, agent retrieval for AI agents, similar items, personalized email, and AI assistants with context-aware recommendations.
How fast is Shaped?
Shaped delivers <50ms query latency for real-time retrieval and ranking. It handles 1,000+ QPS and scales to 5B+ documents with 99.95% uptime SLA.
What data connectors does Shaped support?
Shaped has 30+ native connectors including data warehouses (Snowflake, BigQuery, Redshift, Databricks), analytics applications (Amplitude, Segment, Rudderstack, Posthog), streaming (Kinesis, Kafka, Pub/Sub), and catalog storage (Postgres, MongoDB, Shopify). It supports both batch and streaming data ingestion without ETL.
How do I get started with Shaped?
You can connect your data, build your first model, and deploy to production in 7 days with no infrastructure required. Day 1 is connecting data, Days 2-5 are building and iterating, and Day 7 is pushing live. $100 free credits are available with no credit card required.
What is ShapedQL?
ShapedQL is a SQL interface for real-time retrieval and ranking in Shaped. It allows you to query with SQL and execute a multi-stage pipeline that includes retrieve, filter, score, and reorder operations - all in one query.
Is Shaped secure and compliant?
Yes, Shaped is SOC 2 Type II certified and GDPR, CCPA, and HIPAA compliant. It is enterprise-ready with enterprise-scale capabilities handling 5B+ documents and 99.95% uptime.
What SDKs and APIs does Shaped offer?
Shaped is available via Python SDK, TypeScript SDK, and MCP (Model Context Protocol). It also has a REST API for production integration and a CLI for quickly creating recommendation and search models.
How much does Shaped cost?
Shaped offers $100 free credits with no credit card required. Pricing starts at $0.03 per answer for agent usage, which is 50x cheaper than traditional agent stacks at $1.50 per answer. The company uses flat-fee monthly pricing based on number of users and forecasted usage, with self-serve and white-glove support tiers. Contact sales for enterprise pricing.