Skip to main content

Tela Mentis

Real-time, temporally-aware, multi-tenant knowledge graphs for AI agents
Built with a Rust core and pluggable architecture

Temporal Knowledge Graph Illustration

What problem are we solving?

AI agents struggle with statelessness - they forget important context between interactions. Tela Mentis provides durable, searchable memory that evolves over time, enabling:

Persistent Memory

Store conversations, facts, and relationships in a structured knowledge graph that persists between sessions and across multiple users.

Temporal Awareness

Track both when facts were true and when they were learned, enabling time-aware reasoning and context retrieval.

Multi-Agent Collaboration

Enable multiple AI agents to share a common knowledge base while maintaining tenant isolation for security and privacy.

Knowledge Extraction with LLMs

Tela Mentis seamlessly integrates with LLMs to extract structured knowledge from unstructured text. This enables AI agents to build and maintain their knowledge graphs automatically.

Unified API across OpenAI, Anthropic, and Google models
Extract entities and relationships with temporal context
Automatic schema alignment and deduplication
// Extract knowledge from conversation
const context = {
  messages: [{ 
    role: "user", 
    content: "Alice started working at Acme Corp in January 2023" 
  }]
};

// Knowledge graph is automatically updated
const result = await telaMentis.extract(tenant, context);
Knowledge Extraction Demo

Pluggable Architecture

Tela Mentis is built with a pluggable architecture that allows you to customize any component while maintaining a consistent core API. This makes it adaptable to any AI application stack.

Tela Mentis Architecture Diagram

Key Capabilities

🧠

Millisecond-Scale Performance

Powered by a high-performance Rust core for millisecond-latency graph operations.

Ultra-fast response times essential for natural AI conversations and real-time decision making.
🔌

Adapt to Any Stack

Pluggable architecture with adapters for storage, transport, and LLMs.

Works with Neo4j, In-Memory, HTTP, gRPC, UDS, OpenAI, Anthropic, and Google Gemini.

Time-Aware Reasoning

Bitemporal knowledge model tracks both when facts were true and when they were recorded.

Enables "as-of" queries, temporal reasoning, and change tracking for advanced agent memory.
🏢

Secure Multi-Tenancy

Enterprise-grade isolation between tenants in a single deployment.

Perfect for SaaS applications, serving multiple customers, or isolating different agent ecosystems.
🛠️

Developer-Friendly Tools

Comprehensive CLI tool (kgctl) for all operations and management tasks.

Easily ingest data from CSV, manage tenants, run queries, and export/import graph data.
🤖

AI-Native Integration

Built for AI agents with first-class LLM extraction pipeline.

Transform unstructured text into structured graph knowledge through seamless LLM integration.