First agent in five minutes.

1. Install the SDK

Pick your language. The SDK is the same shape across all three.

# TypeScript / Node
{{TBD-install-command}}  # e.g. npm i @heroa/sdk

# Python
{{TBD-install-command}}  # e.g. pip install heroa

# Go
{{TBD-install-command}}  # e.g. go get github.com/heroa/heroa-go

2. Authenticate

Create a token from the dashboard and export it. All three SDKs read the same env var.

# one-off login flow
{{TBD-auth-flow}}   # e.g. heroa login

# or export a token directly
export HEROA_API_KEY=heroa_sk_...

3. Pick a region

Five regions are available. See regions for residency and latency details.

export HEROA_REGION=us-east

4. Start from a template

Browse twelve templates. Pass the template id to the SDK.

import { heroa } from "@heroa/sdk";

const agent = heroa.agent({
  template: "research-agent",
  region: process.env.HEROA_REGION
});

5. Run the agent

const result = await agent.run({
  input: "Summarize the last six months of Canadian healthcare AI regulation."
});

process.stdout.write(JSON.stringify(result, null, 2));

Expected output, abridged:

{
  "status": "complete",
  "duration_ms": 184502,
  "region": "us-east",
  "summary": "...",
  "sources": [ ... ]
}

6. Let the agent deploy

If the task warrants fan-out, the agent can call heroa.deploy() on its own. See the primitive reference.

7. Inspect in the dashboard

Every run shows up in the dashboard with its compute-minutes, prompt-cache credits applied, and the resulting output artifact.