Connect simulation practitioners, share calibrated environments, and orchestrate agentic workflows across teams — with a single, glass-clear platform.
Trusted by simulation teams at
From environment authoring to agent evaluation, sim-ai.net unifies the simulation lifecycle so your team can ship agents that actually behave.
Connect agents, environments and evaluators as a living graph. Re-route traffic, swap models and replay episodes without touching client code.
Author once, calibrate against real-world telemetry, then share within the network. Every environment ships with provenance and fidelity scores.
Run head-to-head evaluations across hundreds of episodes. Slice by scenario, behavior tree, or reward shape — in seconds, not sprints.
A network registry of community and private environments. Cite, fork, and remix — with cryptographic provenance and granular access.
Stream metrics, traces and tool-calls from every running episode. Drill into a single decision step or zoom out to fleet-wide health.
SOC2 Type II, SAML SSO, BYOK, audit logs and a self-hostable control plane. Your simulations never leave your perimeter unless you allow it.
One client, one network, one mental model. Spin up an environment, attach an agent, and observe the run from any language. The orchestrator handles topology, retries, and snapshotting.
# Connect an agent to a calibrated environment
from simai import Network, Agent, Episode
net = Network.connect("prod")
env = net.environments.get("warehouse-v3")
with net.episode(env, seed=42) as ep:
agent = Agent("my-policy:latest")
for step in ep:
action = agent.act(step.obs)
step.commit(action)
ep.publish("team/warehouse")// Connect an agent to a calibrated environment
import { Network, Agent } from "@simai/sdk";
const net = await Network.connect("prod");
const env = await net.environments.get("warehouse-v3");
await net.episode(env, { seed: 42 }, async (ep) => {
const agent = new Agent("my-policy:latest");
for await (const step of ep) {
const action = await agent.act(step.obs);
await step.commit(action);
}
await ep.publish("team/warehouse");
});# Launch an episode through the REST surface
curl https://api.sim-ai.net/v1/episodes \
-H "Authorization: Bearer $SIMAI_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"environment": "warehouse-v3",
"agent": "my-policy:latest",
"seed": 42,
"publish_to": "team/warehouse"
}'Every team on sim-ai.net contributes to a shared substrate of environments, evaluators, and agent variants — with private rooms when you need them.
Browse a curated registry of simulation environments and agent recipes published by your peers.
Snap environments, agents and evaluators together as a graph. Compose first, configure later.
Pipe in real telemetry to close the sim-to-real gap. Track fidelity scores over time.
Promote variants through staging rings with policy gates and reproducible attestations.
Transparent, predictable plans. Pay for the orchestration time you actually use — never the seats sitting idle.
For solo simulation tinkerers and weekend agents.
For simulation teams shipping real agents.
For platforms with regulated, mission-critical sim.