Jettson Docs

Give your AI agent its own computer. The developer reference for Jettson — what's on an agent's machine, how to bring one online, and the SDKs.

Every AI agent on Jettson gets its own persistent Linux computer — a filesystem, memory, browser, shell, and outbound network it owns and operates across runs. You write the task; the agent installs what it needs, browses the web, does the work, and remembers what it learned.

Start here

  • Quickstart — sign up, get an API key, and an agent comes online in under five minutes.
  • Your first agent — a ten-line example you can copy, run, and modify.
  • Browse examples — three production-grade agents (Stripe triage, PR review, customer research) you can fork and ship.

What an agent gets

When you start an agent it boots its own persistent Linux environment. That environment has a workspace filesystem, a reasoning model, long-term memory, a browser, and outbound network access — the same things a developer would expect on any new box, except this one is owned by the agent and survives across runs.

The agent reads the task, plans, installs whatever it needs, calls out to the web, writes to memory, and returns a structured final_result your application reads. The runtime stays out of the way.

How it works

text
                    ┌──────────────────────────────┐
                    │     An agent comes online    │
                    │      POST /api/v1/agents     │
                    └──────────────┬───────────────┘
                                   │
                    ┌──────────────▼───────────────┐
                    │     The agent's computer     │
                    │      persistent Linux        │
                    └──────────────┬───────────────┘
                                   │
                ┌──────────────────┼──────────────────┐
                │                  │                  │
        ┌───────▼──────┐   ┌───────▼──────┐   ┌──────▼───────┐
        │  Reasoning   │   │  Long-term   │   │   Browser,   │
        │    model     │   │    memory    │   │ files, shell │
        └──────────────┘   └──────────────┘   └──────────────┘
                                   │
                    ┌──────────────▼───────────────┐
                    │     The agent reports back   │
                    └──────────────────────────────┘

The agent plans, calls into its environment, reasons over results, writes important learnings to memory, and returns a structured final_result your application reads.

What agents are doing today

  • Triaging and replying to customer events
  • Reviewing pull requests and learning a team's style over time
  • Researching companies from a single URL and caching the result
  • Running web research that compounds across runs
  • Answering internal questions without standing up a vector database
  • Doing background work for an hour and reporting back

Need help?