Announcing Agentic Postgres: The first database built for agents.
Agents are the New Developer
In a landscape where artificial intelligence is increasingly shaping the software development process, it has been reported that 80% of Claude Code was written by AI, and over a quarter of all new code at Google was generated by AI just a year ago. This trend indicates that within the next 12 months, AI will likely dominate the coding landscape. As agents—intelligent systems that operate differently from humans—emerge, there is a pressing need for software development tools to adapt accordingly. A new kind of database, tailored specifically for the unique behaviors of agents, is essential.
At Tiger, our decade-long dedication to database innovation has led us to recognize the arrival of agents as a pivotal moment in computing. We understood that this new era would require a distinct data infrastructure—one that maintains power while minimizing complexity, designed for a new class of user.
So, how do agents operate?
- Agents don’t click; they call.
- Agents don’t remember; they retrieve.
- Agents can download expertise to become experts.
- Agents can parallelize effortlessly, functioning like a multi-threaded team.
- Agents require a safe sandbox for experimentation.
- Agents can also stress-test your infrastructure and budget if not managed properly.
With these behaviors in mind, we embarked on a year-long journey, involving multiple teams working in parallel, gathering extensive engineering insights and user feedback. The result of this endeavor is something we are proud to unveil.
Introducing Agentic Postgres
Today, we are excited to introduce Agentic Postgres, the first database meticulously crafted for agents. This innovative platform includes:
The best database MCP server ever built
Agentic Postgres features a new MCP server that empowers agents not only to interact with the database but also to understand how to utilize it effectively. Drawing from over ten years of experience with Postgres, we have integrated built-in master prompts that provide agents with safe, structured access to the database, facilitating high-level tasks such as schema design, query tuning, and migrations. Additionally, the MCP server supports native full-text and semantic search across Postgres documentation, allowing agents to retrieve context swiftly as they work.
> I want to create a personal assistant app. Please create a free service on Tiger. Then using Postgres best practices, describe the schema you would create.Native search and retrieval
Agentic Postgres is equipped with native full-text and semantic search capabilities, seamlessly integrated into the database. Enhancements to our existing extension, pgvectorscale, have resulted in higher-throughput indexing, improved recall, and reduced latency, making it superior to pgvector. Our latest extension, pg_textsearch, implements BM25 for modern ranked keyword searches, optimized for hybrid AI applications. The current preview release utilizes an in-memory structure for rapid writes and queries, with future iterations set to introduce disk-based segments featuring compression and BlockMax WAND optimization.
> Using service qilk2gqjuz, analyze user feedback with hybrid search (combining text search and semantic search). Group similar feedback by theme and show counts for each theme, using an ascii bar chart. First, look at the pg_textsearch (BM25) and pgvectorscale documentation in the Tiger docs to get the proper syntax, and then use those extensions.Fast, zero-copy forks
At the heart of Agentic Postgres lies a new copy-on-write block storage layer that enables instantaneous database forking. Each agent can create its own isolated environment, replicating production data in seconds without incurring additional data costs. These lightweight forks are ideal for experiments, benchmarks, and migrations that can run concurrently and safely.
> Please create a fork of gf868h9j1y using the last snapshot, and then test 3 different indexes that we should create to speed up performance, then delete the fork, and report back on your findings. Before you start run “tiger service fork --help” and “tiger service delete --help” to get the right syntax. Use MCP over psql, using the password from the local keychain.New CLI and free tier
We have also developed a new command-line interface (CLI) that simplifies the exploration, forking, and building processes with Agentic Postgres. To facilitate immediate engagement, we are launching a free tier, allowing every developer and agent to dive in right away. Getting started is as simple as executing three commands in your terminal:
# 3 commands to install the Tiger CLI and MCP. That's it!
$ curl -fsSL https://cli.tigerdata.com | sh
$ tiger auth login
$ tiger mcp installAfter that, simply instruct your agent to initiate a new free service using MCP, or execute tiger create service from the command line to commence your journey.
Powered by Fluid Storage
Agentic Postgres operates on Fluid Storage, our newly developed distributed storage layer. Fluid Storage employs a disaggregated architecture of a horizontally scalable distributed block store utilizing local NVMe storage, complemented by a storage proxy layer that exposes copy-on-write volumes and a user-space block device driver. This architecture allows Fluid Storage to present itself as a local disk to Postgres while scaling like a cloud service.
Consequently, Fluid Storage provides instant forks, snapshots, and automatic scaling—both upwards and downwards—without downtime or over-provisioning. In benchmark tests, a single volume has demonstrated the capacity to sustain throughput exceeding 100,000 IOPS, all while retaining the elasticity and copy-on-write capabilities of Fluid. All free services on Tiger Cloud currently operate on Fluid Storage, enabling every developer to experience its performance and flexibility firsthand. This launch marks just the beginning; we will delve deeper into each feature—MCP, pg_textsearch, pgvectorscale, forkable databases, Fluid Storage, CLI, and the free tier—in the coming days and weeks.
Built for Agents and Developers
Agentic Postgres is designed with agents in mind, allowing developers to concentrate on higher-level challenges. Our experience in building with and for agents has revealed a fundamental truth: agents are not here to replace us; they are here to elevate us. By taking on repetitive and mechanical tasks, agents enable us to focus on what truly matters: architecture, design, creativity, and impact. They enhance our speed and intelligence, empowering us to undertake more ambitious projects than we could alone.
The prevailing myth suggests that AI will supplant developers. In reality, it is the developers who collaborate with agents who will thrive, while those who do not will be left behind. Agentic Postgres is tailored for developers eager to build with AI—those who prioritize functional applications over temporary demos, and who desire an engineering experience rather than mere experimentation. Today’s launch is merely the first step; while there may be some rough edges, we welcome your input in refining them. Anticipate further developments—both significant and subtle—in the weeks, months, and years to come.
Agents are the new developers, and Agentic Postgres is their new playground.
Built for Agents. Designed to Elevate Developers.
Let’s embark on this journey together. Get started today:
$ curl -fsSL https://cli.tigerdata.com | sh