Tiger Data today introduced a managed PostgreSQL database service designed specifically for AI agents, saying conventional database architectures are poorly suited to a future in which software is increasingly built and operated by autonomous agents.

The New York-based company, whose business name is Timescale Inc., said the new Ghost service addresses a growing need for infrastructure that supports large-scale experimentation by coding and workflow agents. The service is generally available today.

Founded 10 years ago, Tiger Data has evolved from its roots building a PostgreSQL extension for time-series data “into a database platform that’s used across 3,000 customers,” said co-founder and Chief Executive Ajay Kulkarni. Ghost is built on PostgreSQL but reimagines how databases are provisioned and consumed by AI systems, he said.

“The idea is that all new applications are being built with coding agents,” Kulkarni said. “We believe the agent is the future packaging of software. Instead of an application, users want outcomes.”

However, agent-driven development creates new infrastructure requirements because agents frequently experiment, create temporary environments and test alternative approaches.

“They don’t always do it safely,” Kulkarni said. “The database layer can get really expensive and dangerous if you let your agents experiment and go wild.”

To address that problem, Ghost allows agents to spin up unlimited databases with “fast forking,” a feature that lets users create an exact, independent copy of a dataset in minutes or seconds, bypassing long copying times. Tiger Data charges based on active compute usage rather than per-database licensing.

“Databases are free,” Kulkarni said. “You could have one database per agent or one database per task. You can create it in seconds.”

Copy-on-write storage

The service is built on Tiger Data’s proprietary Fluid Storage technology, a copy-on-write storage layer that allows multiple database instances to share underlying data blocks. Instead of paying for data copies, users only pay for data that changes.

Kulkarni compared the architecture to pointers that reference common storage until modifications occur. “If one database starts diverging and writes new blocks to disk, it only has to keep track of the new blocks,” he said.

Tiger Data is positioning Ghost as an evolution of PostgreSQL to modernize for agents rather than an alternative to it.

The company also emphasizes compatibility with the broader PostgreSQL ecosystem. Ghost supports popular extensions including TimescaleDB, pgvector and PostGIS.

Early adopters are already using Ghost in ways Tiger Data didn’t initially anticipate, Kulkarni said. “It’s kind of serving as a scratch pad for agents,” he said.

He described one case in which the platform was used to analyze operational data by having an AI assistant create a temporary database, load a dataset, perform analysis and then discard the environment afterward.

Despite the rise of new AI-focused data infrastructure like vector databases, Kulkarni expects PostgreSQL to remain at the center of AI development.

“I would be surprised if the database for agents is not Postgres,” he said. “Postgres is reliable, it’s flexible, it’s extensible. It also has a huge ecosystem and community.”

Tiger Data has 200 employees in 25 countries. The company has raised $180 million in funding.

Image: SiliconANGLE/Microsoft Designer