How Stripe Built Kai on Deep Agents in 1 Week
What happened
Stripe created an internal AI assistant called Kai to help its employees with productivity tasks. The development team built Kai in one week using the Deep Agents harness and the LangChain/LangGraph stack. Kai connects with company tools such as Slack and Google Suite, allowing employees to synthesize information and collaborate. The assistant is available to all Stripe staff and is tailored to the company's workflows and data systems.
Key details
- Kai was built in one week by Stripe's AI platform team.
- The system is based on Deep Agents, an open-source agent harness from LangChain.
- Kai integrates with Slack, Google Suite, and Stripe's internal data warehouse.
- Over 1,000 skills from more than 100 teams are available in Kai's library.
- Kai uses middleware for file management, code execution in a sandbox, and summarization to handle long sessions.
Why it matters
Stripe employees now have access to a company-specific AI assistant that streamlines tasks like data analysis, document drafting, and cross-team collaboration. This reduces the technical barriers for non-engineers to use AI tools, allowing more staff to benefit from automation and data synthesis without needing specialized training. The rapid deployment means Stripe can quickly adapt and scale AI-powered workflows across its organization.