Stack Overflow for Agents triples signups after launching ChatGPT plugin

A fast-built product finds its audience

Two engineers spent just a few weeks turning an idea into a working product. Stack Overflow for Agents was born inside Codex, OpenAI's tool that helps developers write code by letting a natural-language plan guide the build process.

Alexandra Lato, vice president of product at Stack Overflow, said the team simply gave Codex a development environment and a plan, then worked alongside it. "So it really made their team feel bigger," she said.

The result was a full-fledged agent product — a tool that lets artificial intelligence agents pull verified knowledge from the many systems where companies store information, like Slack, Google Docs, and internal databases.

One plugin changes everything

Once the product was ready, the team needed a way to put it in front of users. They built a ChatGPT plugin that lets people discover and set up Stack Overflow for Agents directly inside ChatGPT, bypassing the longer setup steps required by the standard Model Context Protocol (MCP).

"Once you have a plugin, it becomes a frictionless setup with ChatGPT," Lato said.

The effect was immediate. Registrations tripled overnight. Agent creation grew by as much as 10 times, according to the company.

Speeding up the hard parts

While the two-engineer push built the agent product quickly, the company's enterprise offering, Stack Internal, still requires connecting to dozens of different systems. That work typically takes far more time. Codex is helping there, too.

On average, building a single connector — the software link that lets Stack Overflow talk to a company's tools — used to take three to six weeks. With Codex, that dropped to four days, a reduction of 87 percent. More recently, the team built one in just a single day. The overall timeline for the enterprise product shrank from a year to roughly three months.

Lato said the platform runs on OpenAI models including GPT-6 Astra and GPT-6 Sol, routing tasks to whichever model fits the reasoning work and cost. She also expressed interest in using OpenAI's Decisions API to automate some of the model-routing choices, hoping to free engineering time for product work instead.

Closing the knowledge loop

Looking ahead, Stack Overflow wants to go beyond pulling information in. The next step is sharing what gets learned so that knowledge gained in one conversation can help the rest of an organization.

"As I'm learning, how do I make sure that the rest of my organization is learning with me?" Lato asked.

For now, the quick wins speak for themselves: a small team, a fast-built product, and one plugin that turned steady interest into a surge.

This story reports on a trend for information only. It is not advice or encouragement to do any of it.

Key facts

• Two engineers shipped Stack Overflow for Agents in a few weeks using Codex
• A ChatGPT plugin tripled registrations overnight
• Agent creation grew up to 10x after the plugin launch
• Connector build time dropped from 3-6 weeks to 4 days on average with Codex
• The platform uses GPT-6 Astra and GPT-6 Sol with model routing

Sources

1. Stack Overflow for Agents triples signups with ChatGPT plugin