> ## Documentation Index
> Fetch the complete documentation index at: https://docs.joystream.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# JoyStream

> The operating system for organizational agents — engineer, operate, and improve the capabilities that run your business.

Every organization has recurring work it must do well: onboarding employees, qualifying leads, closing the books. Those are its capabilities.

JoyStream turns those capabilities into software your organization can inspect, approve, run, and improve.

## The problem it solves

Generating an agent is getting easy. Trusting one is not.

Making an agent you can put into production still takes an engineer — someone to audit the logic, wire up a sandbox, write evaluation loops, and watch what it touches. So the people who own the work can't safely build and evolve it themselves, and organizations are left choosing between building nothing and shipping something they hope won't break anything downstream.

JoyStream gives the person who owns the capability engineering evidence, without requiring a developer. You describe what you want. JoyStream turns that intent into a readable specification, a reviewable plan of the work it will do, and a safe run before it touches anything real.

By the time you say "I want to deploy this," you have the evidence to justify it. Deployment stops being a leap of faith and becomes a review of evidence.

## Start here

<Columns cols={2}>
  <Card title="Install the CLI" href="/installation" icon="terminal">
    Get `jstm` on your machine and sign in.
  </Card>

  <Card title="Quickstart" href="/quickstart" icon="rocket">
    Draft, build, and dry-run your first agent.
  </Card>

  <Card title="How JoyStream thinks" href="/concepts/overview" icon="compass">
    The belief, the category, and the primary object.
  </Card>

  <Card title="CLI reference" href="/cli/overview" icon="book">
    Every command, flag, and exit code.
  </Card>
</Columns>

## The shape of the work

```mermaid theme={null}
flowchart LR
    A[Plan] --> B[Build] --> C[Conform] --> D[Dry Run] --> E[Pilot] --> F[Live]
    F --> G[Improve]
    G -.-> A
```

You plan an agent in conversation, build it into an immutable versioned package, prove the package meets the spec, run it with side effects held back, then promote it — deliberately — to a limited pilot and on to production. Every run leaves behind evidence, tickets, and memory that make the next version better.

Nothing reaches production automatically. People decide when the evidence is enough.

## What makes an agent trustworthy

A workflow is bounded but brittle: rigid steps that shatter when reality shifts. A vibe agent is intelligent but unbounded: it improvises, often outside the guardrails you intended.

A JoyStream agent is both bounded and intelligent. The capability is engineered as a plan, and the agent reasons through the messy middle that scripts can't handle — always within the intent and boundaries that were approved.

Think of it as a route plan. You approve the journey before the car leaves the driveway. The driver still chooses the best road to each destination.

## Capabilities, not agents

An agent trapped on one person's laptop is a liability. When they leave, the capability leaves with them.

Real organizational capabilities were never one person's to begin with. Lead qualification, onboarding, customer escalation, finance — these are owned by teams.

The capability is the asset. The agent is one implementation of it. JoyStream is where capabilities are owned, operated, and improved together — carrying their specification, the conversations that shaped them, the evidence that earned them production, every execution, and the memory they accumulate.
