Edge deployment in 42 regions
Each request goes to the nearest healthy region. If one region fails, traffic moves in under 200 ms.
Now running in 42 regions
Strandsagents Cloud runs your agents close to your users. Tracing, secrets, guardrails, and autoscaling come built in, so you write the agent and skip the infrastructure.
Platform
You bring the model and the tools. Strandsagents Cloud handles where the agent runs, how it scales, and what you can see afterward.
Each request goes to the nearest healthy region. If one region fails, traffic moves in under 200 ms.
Store API keys once. Agents read them at runtime and never write them to logs or images.
Idle agents cost nothing. Replicas start in under a second when traffic arrives.
Each model call, tool call, and token count is traced by default. Open any run and see why the agent chose a tool.
Redact personal data, block unsafe tool inputs, and cap spend per user before a request reaches a model.
Send a small share of traffic to a new version first. Roll back in one command if error rates rise.
How it works
Three steps. No Dockerfile, no YAML, and no cluster to size.
Add the package to your project with one command.
pip install strandsagents
Define the model, the tools, and the guardrails in plain Python.
agent.py
Run one command. The agent is live in every region within seconds.
strandsagents deploy
Developer experience
Tools are ordinary functions with type hints. Strandsagents Cloud reads the signatures, builds the tool schema, and wires up tracing.
# agent.py
from strandsagents import Agent, tool
# Tools are plain Python functions.
@tool
def lookup_order(order_id: str) -> dict:
"""Fetch an order from the warehouse."""
return warehouse.get(order_id)
agent = Agent(
model="claude-sonnet",
tools=[lookup_order],
guardrails=["pii-redaction"],
max_spend_per_user=0.50,
)
agent.deploy(regions="auto", min_replicas=0)
Pricing
Every plan includes tracing, secrets, and all 42 regions.
For side projects and experiments.
For teams shipping agents to customers.
For regulated workloads and large fleets.
Deploy the demo agent in one command and open its traces in the console.