Policy-enforced memory management for AI

Store, access, and manage user memories with built-in privacy controls. Perfect for building personalized AI assistants that respect user preferences and constraints.

Use with your stack

We offer client and server libraries in everything from Python and JavaScript to Go and Ruby.

See libraries →

Policy enforcement

Built-in policy matrix ensures memories are only accessed for allowed purposes. Fail-closed by default.

View docs →

Deterministic merging

Automatically merge multiple memories with deterministic results and confidence scores.

Learn more →

Revocation tokens

Every read operation returns a revocation token. Users can revoke access at any time.

Explore →

Store user preferences

Save user likes and dislikes with automatic normalization and deduplication.

The Memory Scope API makes it easy to store and organize user preferences across different domains. When you create a memory, the API automatically normalizes and deduplicates values, ensuring consistency across your application. Each memory is scoped to a specific purpose and can have a configurable time-to-live.

Automatic normalization and deduplication
Domain-based organization
Configurable TTL for automatic expiration
Support for structured JSON data

Read with policy enforcement

Access memories with purpose-based access control. Policy violations are automatically blocked.

Every read operation requires a purpose, and the API enforces a policy matrix to ensure memories are only accessed for allowed purposes. The policy system is fail-closed by default, meaning any purpose not explicitly allowed is automatically denied. This ensures your users' data is protected even if you make a mistake in your code.

Purpose-based access control
Fail-closed policy enforcement
Automatic memory merging
Confidence scores for merged results

Policy denial example

See how the API blocks unauthorized access attempts. Fail-closed by default.

When a read operation requests access for a purpose that's not allowed by the policy matrix, the API immediately returns a 403 Forbidden response. This fail-closed approach ensures that user data is protected by default, even if your application code has bugs or security vulnerabilities. The error response includes clear details about why the access was denied.

Immediate denial of unauthorized access
Clear error messages
No data leakage on policy violations
Audit logging of all denials

Revoke access

Users can revoke access at any time using the revocation token from a read operation.

Every read operation returns a revocation token that can be used to immediately revoke access to the data that was read. This gives users complete control over their data and allows them to withdraw consent at any time. Once revoked, subsequent read operations for the same scope and domain will return empty results until new memories are created.

Revocation tokens with every read
Immediate access revocation
User-controlled data access
Privacy-first design
request.json1 / 4
POST /memory { "user_id": "user123", "scope": "preferences", "domain": "food", "source": "explicit_user_input", "ttl_days": 30, "value_json": { "likes": ["pizza", "sushi"], "dislikes": ["broccoli"] } }
TERMINALoutput

Use Cases

Power your AI applications with secure memory management

Personalized AI Assistants
Store user preferences and constraints to create truly personalized AI experiences.
Recommendation Systems
Leverage user preferences and constraints to power intelligent recommendation engines.
Content Generation
Generate personalized content based on user communication preferences and style.
Scheduling Applications
Respect user schedule constraints and availability windows for smart scheduling.
E-commerce Personalization
Remember shopping preferences, past purchases, and product interests for tailored shopping experiences.
Customer Support
Maintain context across conversations to provide consistent and personalized support interactions.
Learning Platforms
Track learning preferences, progress, and adapt content delivery based on user learning styles.
Health & Wellness Apps
Store health preferences, dietary restrictions, and wellness goals while maintaining privacy.
Financial Services
Remember financial preferences, risk tolerance, and investment goals for personalized advice.

Ready to get started?

Join developers building the next generation of AI applications