Tiger Data sees agents as the new developer

Ajay Kulkarni, CEO & co-founder of Tiger Data (formerly Timescale), remembers the first time he heard the word "agentic"—two years ago, and he thought it was a joke. Then he used Claude Code for the first time this year. He saw it write code, commit to GitHub, deploy. He couldn't sleep that night. "I could build anything," he kept thinking.

Tiger Data is a database company. They started with time series data—sensor data from IoT, crypto pricing—built on Postgres. But they've evolved with Postgres's rise to become the de facto database. Now they're rebuilding for a world where 80% of Claude Code is written by Claude, and customers are reporting 70% of their code comes from agents.

The interface is changing

Agents don't click, they call. They don't navigate UIs—they call APIs. This means the surface area of developer tools needs to change. Tiger Data has spent a lot of time thinking about MCP server design, which Ajay frames as "affordances for agents."

In the design world, affordances tell you what you can do—a door handle tells you the door opens. MCP tools are affordances for AI agents. You want to give them enough capabilities to be useful, but not so many that they get confused. You want building blocks that can be chained together.

Speed matters differently now

For databases serving agents, speed has new dimensions. Agents parallelize. They experiment. They spin up sandboxes. You need databases that boot quickly, fork quickly, and create isolated environments cheaply.

Agents also don't remember, they retrieve. Context is loaded on demand, which means native search becomes critical. The database needs to help the agent find the information it needs when it needs it.

60% adoption in three weeks

Tiger Data built an internal Slack bot called Eon that can query Salesforce, call logs, usage data, and Slack transcripts. Within three weeks, 60% of the company was using it daily. They built it themselves rather than using a commercial product partly because if you're building for agents, you should probably be building agents.

The broader transformation at the company mirrors what they're hearing from customers. People are doing more with the same headcount. Engineers are contributing code who haven't coded in years. Sales, design, product—everyone is finding their own use cases and sharing them.

AI is the new UI

Ajay quotes Naval Ravikant: "AI is the new UI." He's normally skeptical of pithy statements, but this one feels right. The rise of SaaS shifted power from the CIO to the end user, which made great UI essential. The rise of AI is shifting power again—deemphasizing the GUI, emphasizing terminals, MCPs, and natural language interfaces.

The web UI won't go away, but it will change. Ajay sees another decade of innovation ahead, at least.

When asked about re:Invent, Ajay describes it as "the bazaar", the central market where you come to see what everyone's selling, how they're messaging, what's changed. He's been coming for six years. Five years ago everyone talked about big data and cloud. Now it's all AI. The evolution is part of the show.

This interview was conducted at AWS re:Invent 2025.