What a model does, not what it says.

Press an option, the way a model in AI Observability Studio calls a tool. The receiver shows what arrived, the decision and the run's log.

SenderStands in for a model in AI Observability Studio

Options

Or start from:

Press one, as the model would

Each press opens a new session, runs it to its end and returns the log. Nothing you type is stored.

Every run here is simulated. Nothing a model does reaches the real world. Here you stand in for the model. In AI Observability Studio a real model is connected to an environment like this one, reads a prompt this site never sees, and chooses.

Live decisionsFrom AI Observability Studio and the API

Listening…

2–5Decisions per environment

Real timeThinking costs time

0Prompts seen here

MCPHow tools reach a model

What a decision shows ↗

Build with it

Decisions, a clock, and nothing else.

Load an environment by its id and open a session: your model gets an MCP address for its tools, and you keep the story. What it sends is carried out against the clock, and the decision and the full log come back.

Read the API guide →

POST /v1/sessions

curl https://www.open-simulations.com/v1/environments/env_k3x9q2m7ab \
  -H "Authorization: Bearer os_live_…"

curl https://www.open-simulations.com/v1/sessions \
  -H "Authorization: Bearer os_live_…" \
  -H "Content-Type: application/json" \
  -d '{"scenarioId":"scn_env-k3x9q2m7ab","version":1,
       "variables":{"deadlineSeconds":90}}'