
Simulate human decisions before they happen
Seldon is an applied research lab building models that simulate human decision-making, reflecting how reasoning, emotion, experience and social influence shape behavior. Its simulation agents carry configurable traits like personality, beliefs, risk tolerance and social influence, and the company reports 87% behavioral fidelity against observed human behavior. The same underlying technology powers three products: Wren for software QA testing, Finch for product and user research, and Rook for government wargaming and simulation.
Agents are shaped by personality, goals and motivations, beliefs and expectations, emotions and attention, environment and context, and experience and memory, which carry into the choices they make.
Change who an agent represents or the situation they face to explore how their choices change, with traits such as risk tolerance, decision speed, loss aversion, novelty seeking and social influence.
Model scenarios like a user abandoning checkout, a critical bug reaching production, or a decision under pressure, then see trajectories, friction points and diverging outcomes.
Follow what agents encounter, the choices they make and what happens next; inspect recorded journeys and evidence behind failures or divergent behavior.
Models are evaluated against observed human behavior, and Seldon reports simulated users made the same choices as real people 87% of the time across the full journey.
Define software QA tests in plain English — task, expected behavior and constraints — and let agents explore workflows beyond scripted tests.
Shared technology adapted to three use cases: software QA testing (Wren), product and user research (Finch), and national security/defense simulation including wargaming (Rook).
Define the task, expected behavior and constraints in plain English, or work with Seldon to ground agents in your user context or operating environment.
Agents with different goals, information and experience explore the situation — exercising software workflows, experiencing a product, or making decisions under uncertainty.
See where trajectories diverge, review findings and recorded journeys, and decide what to investigate or change before problems reach production.
Wren (software QA testing) is priced at $3 per completed test run in USD, with model usage and browser execution included, no token charges or seat fees, and the first 10 completed test runs free per workspace. Wren Enterprise offers volume pricing for larger teams. Finch (product research) is enterprise-only with custom pricing. Rook (national security and defense) is government-only with custom
$3 per completed test run
Software QA testing. Includes model usage, browser execution, findings and recorded evidence. No token charges or seat fees. First 10 completed test runs free (one allowance per workspace).
Custom
Volume pricing and enterprise requirements for larger teams.
Custom (Enterprise)
Product and user research. Work with Seldon to scope research, ground agents in user context, and evaluate agent behavior.
Custom
National security and defense simulation, government organizations only, via custom engagements.
Seldon is an applied research lab building models that reflect how reasoning, emotion, experience and social influence shape human decisions, with the ambition of establishing a new standard for modeling human behavior. Its products apply this simulation technology to software QA testing (Wren), product research (Finch) and national security/defense (Rook).
Wren costs $3 per completed test run in USD, with model usage and browser execution included, no token charges or seat fees, and the first 10 completed test runs free per workspace. Finch and Wren Enterprise use custom pricing, and Rook is available to governments through custom engagements.
One test case executed once in one browser and environment. A suite of 10 cases executed once is 10 runs ($30 before the free allowance), and running the same case again counts as another run. A completed test that identifies a bug in your application counts as a run.
The page does not describe Seldon or its products as open source; products are accessed through sign-up, demos or custom engagements with the company.
Yes — your first 10 completed test runs are free, with one shared allowance per workspace. Sign up, create a workspace, and try a workflow your team already tests.
If Wren cannot complete a run because of an agent or platform failure, it does not count as a charged run or use your free allowance.
Rook is available only to government organizations, through custom engagements covering your scenario, operating environment and deployment requirements.
Seldon reports 87% behavioral fidelity: when comparing simulated users with real people across the full journey, they made the same choices 87% of the time. The company evaluates models against observed human behavior to identify where simulation succeeds and where limitations remain.