WorkOS: From Evaluating Mastra to Teaching It
When Mastra went viral on GitHub in February, Nick Nisi was watching.
Nick is a host on popular developer podcasts like JS Party and Software Engineering Daily. Along with Zack Proser, he leads developer relations at WorkOS, an auth provider for companies like OpenAI and Plaid that recently crossed $20m in annual revenue
One day, Nick was trying to teach a teammate a concept around agents, and found himself spinning up Mastra to use the local dev playground. "I went from wondering how I would demonstrate this complex AI pipeline to suddenly having a professional interface that just worked," he recalls.
One part of Nick's job is creating proofs of concept, and Mastra solved an immediate problem. Post-Mastra, he could stay focused on business logic, without worrying about interface design, while still ending up with a polished interface ready to present ideas to the team.
He demoed Mastra at the WorkOS all-hands in April, the response was immediate. Within days, multiple projects had spun up across the org.
"Mastra's TypeScript-native approach really resonated with our team," explains Zack. "We could see immediately how it would fit into our existing development workflow."
Mastra in Production
Jacobia Johnson is a software engineer at WorkOS, and in May and June after Nick's demo, she became one of the first engineers to launch Mastra workflows into production. Her role is centered around GTM productivity.
After Nick's demo, Jacobia built a number of Mastra workflows. One tool takes unformatted text including prospect information, identifies companies and people, runs it through enrichment processes, and outputs a cleaned, formatted table. Another scrapes website URLs, filters for target personas, and returns formatted tables. For example, you could pass in a conference website and get speaker data filtered for WorkOS ICPs.
Mastra's flexibility was crucial. "There are enrichment tools out there, but they're often too rigid," Johnson notes. "We get lists in different formats. Before Mastra, we'd have to set up different workflows every time."
Jacobia uses Mastra as a custom layer on top of Cargo, their enrichment provider. "Cargo lets you expose tools as an MCP server, so I connected our agents to it and they can decide which enrichment tools to use."
She also appreciates the easy deployment to Mastra Cloud. "Honestly, I'm not sure how I would have built these agents without [Mastra]," Johnson says. "The playground has also been really nice for easily testing different flows."
Engaging the community
The relationship has extended beyond pure technical work.
Nick and Zack's led a 2-hour hands-on Mastra workshop at the AI Engineer World's Fair in San Francisco. The session, attended by over 100 engineers, focused on building agentic systems using Mastra.
The playground let Nick and Zack build in real time and demonstrate capabilities without getting bogged down in setup, and built a meme generator with MCP integrations.
"The room was packed, and the energy was high," Zack noted afterwards. "We walked folks through building a full multi-step agent workflow, and it was great to see people rolling up their sleeves and diving in."
It's a microcosm of a working relationship showing how AI-driven tools work together.
One evening in May a couple weeks before, I demoed at WorkOS's MCP Night at the San Francisco Exploratorium, an event that drew hundreds of engineers, founders, and researchers.
The evening after Nick and Zack's talk, I was at dinner with WorkOS CEO Michael Grinich, swapping growth strategies (when I explained our plan to distribute thousands of copies of Principles of Building AI Agents every month, he challenged me to double it).
The agent movement moves forward -- and there's dozens or hundreds of agentic use-cases inside most organizations waiting to be discovered.