Agents are emerging as the new users of databases, diverging from the linear execution typical of traditional applications. These agents will soon be integrated into every workflow, branching out, exploring, and adapting in parallel. To support this evolution, databases must provide infinite, secure copies of data for experimentation, evaluation, and continuous learning. For developers, this translates to the necessity for reasoning agents to instantly fork environments for workflow testing, while data agents require reproducible forks to analyze live information concurrently.
Current databases and makeshift stacks fall short of meeting these future demands. The reliance on costly replicas, fragile pipelines, and disjointed tools hampers efficiency and undermines production reliability. Without a database specifically designed for agents, the entire field risks stagnation.
Introducing Agentic Postgres
Agentic Postgres addresses these challenges. Now available on Tiger, Tiger Data’s fully managed Postgres cloud, it introduces a groundbreaking concept: forkable infrastructure. For the first time, developers and agents can create instant, copy-on-write branches of both databases and volumes, facilitating safe parallel experiments without incurring excessive costs. In addition to this innovation, Agentic Postgres incorporates three new primitives into Postgres: an interface for agent control, hybrid search capabilities, and persistent memory for state management.
“We defined the database for IoT. With Agentic Postgres, we are defining the database for agents: a new kind of user that requires infinite, reproducible environments to branch, explore, and adapt,” stated Ajay Kulkarni, Co-Founder and CEO of Tiger Data. “Tiger already powers Fortune 500 companies in IoT, Web3, and AI, proving its readiness for production at scale.”
How Agentic Postgres Delivers Essential Primitives
Agentic Postgres delivers these primitives within the Tiger ecosystem, starting with its innovative forkable infrastructure.
Forkable Infrastructure
Forkable infrastructure combines two essential capabilities that together unlock safe, instant parallelism for agent workloads:
- Forkable Databases: This feature allows for the creation of instant, zero-copy branches of Postgres itself, including schema, tables, and rows. Developers can safely test schemas, simulate new logic, or debug without impacting production. Powered by copy-on-write mechanics, these forks are lightweight and provisioned in seconds.
- Forkable Volumes: This extends beyond the database to encompass the entire environment, including attached storage, embeddings, indexes, and artifacts. Each fork serves as a complete, reproducible snapshot, enabling agents to conduct experiments or parallel tasks as if they were operating in a production environment.
Together, forkable databases and volumes enable the rapid creation of numerous environments today, with an architecture designed to scale toward effectively infinite parallelism. Forks incur charges only for changes made, rather than full copies, providing developers with the speed, safety, and cost-effectiveness they require. (Public Preview)
Interface, Search, and Memory
In addition to forkability, Agentic Postgres integrates foundational primitives directly into Postgres:
- Interface: A control plane accessible through REST APIs, CLI, and MCP endpoints. (Available Today)
- Search: Hybrid retrieval capabilities within Postgres. Vector search with pgvectorscale is currently available, while BM25 keyword search is in public preview.
- Memory: Persistent context for conversations, preferences, and shared state, available via APIs and MCP endpoints. (Public Preview)
The combination of interface, search, and memory forms the foundation, while forkable infrastructure represents the breakthrough. Together, they position Tiger as the Postgres database tailored for agents.
Advantages Over Fragmented or Proprietary Tools
Many existing approaches to agent infrastructure hinder developer productivity. Teams often resort to stitching together vector stores, memory services, and orchestration layers, resulting in brittle, costly, and difficult-to-secure pipelines. While some newer systems offer forkable databases, they frequently lack forkable volumes, leaving developers unable to replicate complete environments. Proprietary stacks can further restrict teams to specific vendors, limiting flexibility.
Agentic Postgres adopts a different approach. It maintains compatibility with Postgres at its core, allowing developers to leverage the ecosystem they already trust, while Tiger’s next-generation storage layer introduces forkable infrastructure that Postgres alone cannot provide. This results in instant, safe, infinite parallelism for agents, all without abandoning Postgres or succumbing to vendor lock-in.
The outcome is a production-ready foundation that merges speed, safety, cost-effectiveness, and Postgres-native simplicity, designed for the agentic era.
Getting Started with Agentic Postgres
Developers can explore Agentic Postgres today through a new free tier that provides hands-on access to all the essential primitives—forkable databases, hybrid search, memory APIs, and MCP integration—at no cost. This allows developers to start in minutes, experience the advantages of forkable infrastructure firsthand, and seamlessly scale into production.
The free tier includes:
- Forkable database and volume, BM25 keyword search, and memory APIs (Public Preview)
- Expanded MCP integration and vector search with pgvectorscale (Available Today)
Having established the database layer for billions of IoT devices, Tiger Data is now focused on creating the database for billions of agents. In this new era, the database is Tiger.
About Tiger Data
Tiger Data is the driving force behind TimescaleDB and Agentic Postgres, dedicated to building modern data infrastructure for agents, devices, and developers. With over 2,000 customers—including OpenSea, HuggingFace, Mistral, Mercor, and Siemens—Tiger powers production workloads across IoT, Web3, and AI. Founded in 2017, Tiger Data has successfully raised over 0 million from prominent investors such as Benchmark, NEA, Redpoint, and Tiger Global.
