TinyTroupe is a Microsoft open-source library for simulating people with large language models. You define artificial personas โ with names, personalities, occupations and quirks โ and let them think, talk and act inside simulated environments, then observe what the group comes up with. The project has about 7,600 stars on GitHub, is released under the MIT license, and was updated in 2026.
The library is built for imagination and insight rather than automation. A TinyPerson agent maintains a mental state, memory and a stream of thoughts, perceives a TinyWorld environment, and interacts with other personas through conversation and action. Marketers use it to pre-test ideas against synthetic focus groups, product teams use it to role-play customers before real research, and writers and game designers use it to explore how characters might behave. Because the personas are prompted LLM agents, experiments are cheap and repeatable: you can rerun the same scenario with twenty different audiences in an afternoon, something no real panel study could offer. TinyTroupe provides the agent scaffolding, environment mechanics and logging so you can focus on designing the simulation.
Free. TinyTroupe is open source under the MIT license; the only cost is the LLM API usage your simulations consume.
Product and marketing researchers, UX teams, game designers and curious developers who want to prototype ideas against simulated audiences before spending on real-world studies.