Already using the Anthropic or OpenAI SDK? Ship your agent to production in one call.

See the quickstart

Quickstart

Sign up, get an API key, and bring your first agent online. Five minutes.

You'll have a working agent — running on its own computer — by the end of this page.

Prerequisites

  • A terminal with curl and jq (any macOS / Linux box; on Windows use WSL).
  • A free Jettson account — no credit card required for the Free tier.

1. Sign up

Open jettson.dev/signup and create an account with your work email. The Free tier gets you 50 agent-hours a month, 1 concurrent agent, and 100 stored memories — plenty for the quickstart.

2. Generate an API key

In the Console, open API Keys in the sidebar and click Create key. Give it a name (local-dev is fine), then copy the value — it's only shown once.

Keys look like:

text
jett_sk_live_a1b2c3d4e5f6g7h8i9j0k1l2m3n4o5p6

3. Export the key

bash
export JETTSON_API_KEY="jett_sk_live_..."

4. Spawn your first agent

bash
curl -X POST https://jettson.dev/api/v1/agents \
  -H "Authorization: Bearer $JETTSON_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "task": "Write a haiku about containers. Then return JSON with the haiku and a one-line interpretation."
  }' | jq

The response comes back immediately with an agent ID:

json
{
  "agent_id": "ag_8cGoiD4ujxjGOrnjC0QsV",
  "name": "Untitled Agent",
  "status": "spawning",
  "task": "Write a haiku about containers...",
  "created_at": "2026-05-14T18:42:00.000Z",
  "model": "jettson-6.0",
  "container": { "region": "iad", "status": "starting" }
}

5. Watch it run

Open https://jettson.dev/console/agents. Your new agent is at the top — click it to see live progress: planning, tool calls (none needed for a haiku), and the final result.

For an active warm pool the whole run takes under five seconds. Cold spawns take 10-15 seconds the first time.

6. Read the result

bash
curl https://jettson.dev/api/v1/agents/ag_8cGoiD4ujxjGOrnjC0QsV \
  -H "Authorization: Bearer $JETTSON_API_KEY" | jq '.final_result'

You'll get back the agent's structured output. Done.

What just happened

  • POST /api/v1/agents brought an agent online with its own Linux environment — workspace, reasoning model, long-term memory, browser, shell, files, HTTP.
  • The agent read your task, decided no tools were needed for a haiku, wrote the answer, and emitted a terminal "agent_completed" event.
  • Its computer tore down. The agent doc has the full progress trail and final result.

Next steps