Playlab

Playlab is a creator platform where educators build their own AI apps instead of choosing from someone else's template. An app is grounded in your own curriculum, rubrics and references rather than general model knowledge, runs on a model you pick (Anthropic, OpenAI, Google, Meta) with the cost visible while you build, and can draw on pre-mapped knowledge graphs for Illustrative Mathematics, OpenSciEd, state standards and OER Commons.

What you can do with it

Build and remix: start from scratch or adapt one from the community gallery (hundreds of apps, student-built ones included). Steer behavior per app — image and voice input, memory and profiles, or Avoid Answer Generation, which guides students toward an answer instead of handing it over. Also available: a document editor with Docx and Drive export, Python execution, math input and image processing. Teachers can review any student conversation, and automated flags can pause one and notify the teacher. Rosters connect through Clever, ClassLink or Google SSO.

Who it is for

K-12 teachers and districts wanting classroom controls and curriculum grounding without building infrastructure; university instructors and instructional designers wanting one custom course tool; and students aged 13 and over (guardian permission under 18) who build apps themselves. It is a 501(c)(3) nonprofit: educator accounts are free in perpetuity, while school and district tiers are paid.

Notes

Access to the builder is gated: the get-started page asks for an email address and a spot in an onboarding workshop, or an invite code. Its compliance claims (FERPA, COPPA, GDPR, CCPA, SOC 2 Type II), the promise that your work never trains outside models and the assignment of output rights to the creator belong in the data processing agreement, not only the marketing page; its capability comparison against MagicSchool and the major chatbots is Playlab's own, dated mid-2026. It grew out of end-user programming work linked to Scratch and the MIT Media Lab.

Connected Concepts

AI Literacy, Learning Design, Generative AI, RAG (Retrieval-Augmented Generation), Human-in-the-Loop, Privacy, Guardrails