--- title: How Stripe Built Kai on Deep Agents in 1 Week | daily.dev description: Stripe built Kai, a company-wide AI productivity agent, in one week using LangChain, LangGraph, and Deep Agents. Kai serves every Stripe employee with... canonical: https://daily.dev/posts/how-stripe-built-kai-on-deep-agents-in-1-week-ydwbivsmd twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: How Stripe Built Kai on Deep Agents in 1 Week | daily.dev og:description: Stripe built Kai, a company-wide AI productivity agent, in one week using LangChain, LangGraph, and Deep Agents. Kai serves every Stripe employee with... og:url: https://daily.dev/posts/how-stripe-built-kai-on-deep-agents-in-1-week-ydwbivsmd og:image: https://api.daily.dev/og/posts/YdWbivSMD.png og:image:alt: How Stripe Built Kai on Deep Agents in 1 Week og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # How Stripe Built Kai on Deep Agents in 1 Week **[LangChain](https://daily.dev/sources/langchain)** · 8 min read · 4 upvotes · 1 comments ## Summary Stripe built Kai, a company-wide AI productivity agent, in one week using LangChain, LangGraph, and Deep Agents. Kai serves every Stripe employee with context-aware assistance for tasks like data synthesis, document drafting, and analytics — connecting to internal data warehouses, Slack, and Google Suite. The architecture uses Deep Agents as a foundational harness handling tool-calling loops, middleware (filesystem, sandbox, summarization), and state management, with a Stripe-specific layer on top. Kai grew from 296 to over 5,000 users in four weeks, now reaching 83% weekly active usage across 60k+ sessions. Business functions like Marketing (95%) and GTM teams (87%) show the highest adoption. Future work includes scaling skill selection beyond 1,000+ skills, governance guardrails, behavioral personalization, and multi-user collaborative sessions. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What middleware components from Deep Agents did Stripe use to build their Kai AI agent so quickly? Stripe used three Deep Agents middleware components: filesystem middleware (a virtual filesystem backed by S3 with a sync-in/sync-out pattern for sandbox execution), sandbox middleware (isolating Python-based analytics and file processing from the agent's own execution to avoid code-execution security risks), and summarization middleware (managing context in long multi-turn sessions with tunable summarization thresholds and models). _See how daily.dev surfaces practical agent-harness patterns for teams building production AI tools._ ### How many users adopted Stripe's internal AI agent Kai and how fast did it grow? Kai grew more than 16x from 296 users to over 5,000 users in roughly four weeks after entering open preview, hitting the team's quarterly adoption target within the first week. Eventually 83% of Stripe employees used it weekly across more than 60,000 sessions, with especially high adoption in marketing (95%) and go-to-market teams (87%). _Track real-world adoption stories like Kai's on daily.dev when evaluating internal AI agent rollouts._ ### How does Stripe prevent security issues when its Kai agent executes LLM-generated code? Stripe keeps the sandbox separate from the agent itself: the agent runs outside the sandbox and calls into it as a tool rather than executing code directly, which keeps execution boundaries clean and avoids a class of security concerns tied to running LLM-generated code inside the same context as the agent. _Developers wiring code execution into agents can compare sandboxing approaches like this on daily.dev._ ## Community discussion Top comments from developers on daily.dev. **@kartiknvj** · 0 upvotes > The 83% weekly-active number is the part that's hard to fake, because internal tools usually crater to single digits once the novelty wears off. Getting from 296 to 5000+ users in four weeks on a one-week build says the agent was solving a real recurring pain. Do you know what the retention story looked like past the four-week mark once the early-adopter curve flattened? --- Tags: [#ai](https://daily.dev/tags/ai), [#ai-agents](https://daily.dev/tags/ai-agents), [#langchain](https://daily.dev/tags/langchain), [#stripe](https://daily.dev/tags/stripe), [#langgraph](https://daily.dev/tags/langgraph) [View this post on daily.dev](https://daily.dev/posts/how-stripe-built-kai-on-deep-agents-in-1-week-ydwbivsmd) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"How Stripe Built Kai on Deep Agents in 1 Week","url":"https://daily.dev/posts/how-stripe-built-kai-on-deep-agents-in-1-week-ydwbivsmd","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/how-stripe-built-kai-on-deep-agents-in-1-week-ydwbivsmd"},"datePublished":"2026-08-03T15:32:25.034Z","dateModified":"2026-09-13T18:42:13.713Z","description":"Stripe built Kai, a company-wide AI productivity agent, in one week using LangChain, LangGraph, and Deep Agents. Kai serves every Stripe employee with...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/b23b0380ab75e9c7222c4475ec1b9e72?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/b23b0380ab75e9c7222c4475ec1b9e72?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"LangChain","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"LangChain","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/0f4f2629e0724a849823f8cd0d913e13","url":"https://daily.dev/sources/langchain"},"commentCount":1,"discussionUrl":"https://daily.dev/posts/how-stripe-built-kai-on-deep-agents-in-1-week-ydwbivsmd","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":4},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"ai,ai-agents,langchain,stripe,langgraph","timeRequired":"PT8M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"LangChain","item":"https://daily.dev/sources/langchain"},{"@type":"ListItem","position":3,"name":"How Stripe Built Kai on Deep Agents in 1 Week"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/how-stripe-built-kai-on-deep-agents-in-1-week-ydwbivsmd","comment":[{"@type":"Comment","text":"The 83% weekly-active number is the part that’s hard to fake, because internal tools usually crater to single digits once the novelty wears off. Getting from 296 to 5000+ users in four weeks on a one-week build says the agent was solving a real recurring pain. Do you know what the retention story looked like past the four-week mark once the early-adopter curve flattened?","datePublished":"2026-08-03T19:36:41.706Z","url":"https://daily.dev/posts/YdWbivSMD#c-mstycW9YP","author":{"@type":"Person","name":"kartik-nvjk","url":"https://daily.dev/kartiknvj","image":"https://media.daily.dev/image/upload/s--3gGgsVCw--/f_auto/v1781456774/avatars/avatar_TvTVeiMdkRCqWUDullFmy?_a=BAMAMiWQ0"}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/how-stripe-built-kai-on-deep-agents-in-1-week-ydwbivsmd#faq","mainEntity":[{"@type":"Question","name":"What middleware components from Deep Agents did Stripe use to build their Kai AI agent so quickly?","acceptedAnswer":{"@type":"Answer","text":"Stripe used three Deep Agents middleware components: filesystem middleware (a virtual filesystem backed by S3 with a sync-in/sync-out pattern for sandbox execution), sandbox middleware (isolating Python-based analytics and file processing from the agent's own execution to avoid code-execution security risks), and summarization middleware (managing context in long multi-turn sessions with tunable summarization thresholds and models). See how daily.dev surfaces practical agent-harness patterns for teams building production AI tools."}},{"@type":"Question","name":"How many users adopted Stripe's internal AI agent Kai and how fast did it grow?","acceptedAnswer":{"@type":"Answer","text":"Kai grew more than 16x from 296 users to over 5,000 users in roughly four weeks after entering open preview, hitting the team's quarterly adoption target within the first week. Eventually 83% of Stripe employees used it weekly across more than 60,000 sessions, with especially high adoption in marketing (95%) and go-to-market teams (87%). Track real-world adoption stories like Kai's on daily.dev when evaluating internal AI agent rollouts."}},{"@type":"Question","name":"How does Stripe prevent security issues when its Kai agent executes LLM-generated code?","acceptedAnswer":{"@type":"Answer","text":"Stripe keeps the sandbox separate from the agent itself: the agent runs outside the sandbox and calls into it as a tool rather than executing code directly, which keeps execution boundaries clean and avoids a class of security concerns tied to running LLM-generated code inside the same context as the agent. Developers wiring code execution into agents can compare sandboxing approaches like this on daily.dev."}}]} ```