Stack Overflow triples its sign-ups thanks to a ChatGPT plugin

Stack Overflow claims an acceleration of its AI agent offerings, citing a tripling of sign-ups linked to a ChatGPT plugin and connectors developed in just a few days with Codex. The company is preparing to automate its model routing and aims to expand its product for knowledge sharing across the organization.

Stack Overflow plans to use the Decisions API to simplify routing

Stack Overflow is currently managing its own model routing and the associated configuration. Alexandra Lato, Vice President of Product, wants to explore the use of the Decisions API to reduce this workload and free up engineering resources. She believes that automating this routing through OpenAI would save time and money, allowing the team to focus on product development rather than managing safeguards. Additionally, Stack Overflow is working to broaden its knowledge product to enable the sharing of learnings, whereas the initial goal was to gather information. The company aims for the knowledge gained during an individual session to also benefit other members of the organization, including those outside the engineering domain. Alexandra Lato expresses her interest in building a product that covers the complete knowledge lifecycle.

Codex reduces development timelines and accelerates connectors

For Stack Internal, Stack Overflow's enterprise product, Alexandra Lato mentions a reduction in time to market from twelve months to three. She specifies that, on average, building a connector took three to six weeks, but with Codex, a connector was completed in four days, representing a maximum reduction of 87%. The company reports that one connector was even built in a single day recently. Codex thus accelerates the connection work of Stack Internal, which needs to connect to all systems where organizations store their knowledge, a task requiring more human involvement than the initial development of Stack Overflow for Agents by two engineers.

The ChatGPT plugin facilitates service discovery and configuration

To ease distribution, the team developed a ChatGPT plugin that allows users to discover and configure Stack Overflow for Agents directly within ChatGPT, with fewer steps than a traditional configuration of the Protocol Context Model. Alexandra Lato believes this format makes configuration frictionless. She attributes the plugin to a tripling of sign-ups overnight and an acceleration in agent creation, which reached up to ten times the previous level during its last check.

A unified knowledge platform, supported by OpenAI models

Stack Overflow recognizes that both people and AI agents need context that is often scattered across Slack, Google Docs, and other systems, making its assembly complex. The company is building a knowledge intelligence platform to unify these sources, transform raw context into verified knowledge, and make it accessible where work is done. The platform's AI relies on OpenAI models such as GPT-6 Astra and GPT-6 Sol, with task routing to different models based on the reasoning required and costs. Internally, Codex assists engineers in development, and a ChatGPT plugin facilitates the discovery and use of Stack Overflow for Agents. The latter extends Stack's intelligence layer to agents and was developed by two engineers who, according to Alexandra Lato, provided Codex with an environment and a plan to build a complete product in just a few weeks. She believes the tool has given the impression of a larger team. Stack Overflow for Agents is described as an initiative of a North American tech startup, with Codex and an API among its products.