Is Your Database Ready for an Army of AI Agents?

Tiger Data founders Mike Freedman and Ajay Kulkarni discuss their new "agentic Postgres," a database designed for AI agents, addressing its unique requirements compared to traditional databases. They argue that agents, unlike humans, require features like safe play environments, parallelization, and efficient context retrieval. Agentic Postgres incorporates innovations like a new storage layer, fluid storage, MCP server, VectorScale, and BM25 full-text search. The conversation explores the evolution of databases, the role of Postgres, and the importance of memory and state management for agents. They also touch on the potential for agents to revolutionize software development and database management, emphasizing the need for guardrails and security measures.

Outlines

Part 1: Introduction, Agentic Ergonomics

Part 2: Database Evolution, Architecture

Part 3: Storage Layer, Technical Innovation

Part 4: MCP Server, Knowledge Injection

Part 5: State, Memory Management

Part 6: Context, Semantic Layers

Part 7: Search, Retrieval Strategies

Part 8: Future Tools, Ecosystem

Part 9: Open Source, Extensions

Sign in to continue reading, translating and more.

Continue

The Data Exchange with Ben Lorica

Part 1: Introduction, Agentic Ergonomics

Introducing Agentic Postgres: A Database Designed for AI Agents

Why Build Agentic Postgres on Top of the Existing Postgres Foundation?

How Agents Differ from Humans: Rethinking Databases for Agentic Ergonomics

Part 2: Database Evolution, Architecture

Operational vs. Analytical Databases: The Role of Postgres in AI Applications

Avoiding Anthropomorphism: Evolving Postgres for Agents vs. Building from Scratch

Ephemeral Databases: Balancing Full-Blown Systems with Task-Specific Needs

Part 3: Storage Layer, Technical Innovation

Fluid Storage: Enabling Fast, Zero-Copy Forking for Agentic Exploration

Accessing the Storage Layer: Network Protocols and the Developer Experience

Co-Designing with Agents: Enhancing Development with Agentic Postgres and Claude

Part 4: MCP Server, Knowledge Injection

Prompt Templates and Skills: Injecting Domain Knowledge into AI Agents

The MCP Server: Teaching Models to Be Better Database Developers

Postgres and Unstructured Data: Operational Databases and the Rise of HTAP

Part 5: State, Memory Management

State and Memory Management: Databases as the Ideal Solution for Agents

EON: Tiger Agents for Work and the Practicalities of Memory Management

Guardrails for Agent Provisioning: Rate Limiting, Security, and Forkability

Part 6: Context, Semantic Layers

Semantic Layers for Operational Databases: Context and Prompt Injection

Postgres as the First Data Warehouse: Fast Forks for Analytics and Data Exploration

Operational Database Complexity: Guiding Agents to the Right Data Slice

Part 7: Search, Retrieval Strategies

The Flywheel Effect: Capturing Agent Learning and Knowledge Sharing

Memory as a Structured Retrieval Problem: Vector Search and Keyword Search

Transparent Memory and Learning: Crud Access and Reinforcement Learning

Part 8: Future Tools, Ecosystem

Data Access and Usage: Balancing Security and Relevance in Agentic Systems

Reinventing Database Tools for Agents: Low-Level Instrumentation and SRE

New Workloads and the Edge: The Focus on Startups and Hobby Projects

Part 9: Open Source, Extensions

Open Source and the Free Tier: Serving Builders with Ephemeral Architectures

Postgres Plugins and Multimodal Data: VectorScale, BM25, and Hybrid Search

Is Your Database Ready for an Army of AI Agents?

The Data Exchange with Ben Lorica

Part 1: Introduction, Agentic Ergonomics

00:03Introducing Agentic Postgres: A Database Designed for AI Agents

Introducing Agentic Postgres: A Database Designed for AI Agents

01:14Why Build Agentic Postgres on Top of the Existing Postgres Foundation?

Why Build Agentic Postgres on Top of the Existing Postgres Foundation?

03:32How Agents Differ from Humans: Rethinking Databases for Agentic Ergonomics

How Agents Differ from Humans: Rethinking Databases for Agentic Ergonomics

Part 2: Database Evolution, Architecture

05:22Operational vs. Analytical Databases: The Role of Postgres in AI Applications

Operational vs. Analytical Databases: The Role of Postgres in AI Applications

08:41Avoiding Anthropomorphism: Evolving Postgres for Agents vs. Building from Scratch

Avoiding Anthropomorphism: Evolving Postgres for Agents vs. Building from Scratch

10:56Ephemeral Databases: Balancing Full-Blown Systems with Task-Specific Needs

Ephemeral Databases: Balancing Full-Blown Systems with Task-Specific Needs

Part 3: Storage Layer, Technical Innovation

12:29Fluid Storage: Enabling Fast, Zero-Copy Forking for Agentic Exploration

Fluid Storage: Enabling Fast, Zero-Copy Forking for Agentic Exploration

15:10Accessing the Storage Layer: Network Protocols and the Developer Experience

Accessing the Storage Layer: Network Protocols and the Developer Experience

17:18Co-Designing with Agents: Enhancing Development with Agentic Postgres and Claude

Co-Designing with Agents: Enhancing Development with Agentic Postgres and Claude

Part 4: MCP Server, Knowledge Injection

19:00Prompt Templates and Skills: Injecting Domain Knowledge into AI Agents

Prompt Templates and Skills: Injecting Domain Knowledge into AI Agents

21:15The MCP Server: Teaching Models to Be Better Database Developers

The MCP Server: Teaching Models to Be Better Database Developers

24:42Postgres and Unstructured Data: Operational Databases and the Rise of HTAP

Postgres and Unstructured Data: Operational Databases and the Rise of HTAP

Part 5: State, Memory Management

26:16State and Memory Management: Databases as the Ideal Solution for Agents

State and Memory Management: Databases as the Ideal Solution for Agents

27:43EON: Tiger Agents for Work and the Practicalities of Memory Management

EON: Tiger Agents for Work and the Practicalities of Memory Management

30:24Guardrails for Agent Provisioning: Rate Limiting, Security, and Forkability

Guardrails for Agent Provisioning: Rate Limiting, Security, and Forkability

Part 6: Context, Semantic Layers

32:13Semantic Layers for Operational Databases: Context and Prompt Injection

Semantic Layers for Operational Databases: Context and Prompt Injection

34:00Postgres as the First Data Warehouse: Fast Forks for Analytics and Data Exploration

Postgres as the First Data Warehouse: Fast Forks for Analytics and Data Exploration

35:32Operational Database Complexity: Guiding Agents to the Right Data Slice

Operational Database Complexity: Guiding Agents to the Right Data Slice

Part 7: Search, Retrieval Strategies

38:32The Flywheel Effect: Capturing Agent Learning and Knowledge Sharing

The Flywheel Effect: Capturing Agent Learning and Knowledge Sharing

40:34Memory as a Structured Retrieval Problem: Vector Search and Keyword Search

Memory as a Structured Retrieval Problem: Vector Search and Keyword Search

42:26Transparent Memory and Learning: Crud Access and Reinforcement Learning

Transparent Memory and Learning: Crud Access and Reinforcement Learning

Part 8: Future Tools, Ecosystem

45:53Data Access and Usage: Balancing Security and Relevance in Agentic Systems

Data Access and Usage: Balancing Security and Relevance in Agentic Systems

47:55Reinventing Database Tools for Agents: Low-Level Instrumentation and SRE

Reinventing Database Tools for Agents: Low-Level Instrumentation and SRE

50:03New Workloads and the Edge: The Focus on Startups and Hobby Projects

New Workloads and the Edge: The Focus on Startups and Hobby Projects

Part 9: Open Source, Extensions

51:27Open Source and the Free Tier: Serving Builders with Ephemeral Architectures

Open Source and the Free Tier: Serving Builders with Ephemeral Architectures

53:07Postgres Plugins and Multimodal Data: VectorScale, BM25, and Hybrid Search

Postgres Plugins and Multimodal Data: VectorScale, BM25, and Hybrid Search