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

# Introduction

> The operating system for your infinite workforce

## The operating system for your infinite workforce

DeepMyst is the operating system for your infinite workforce. Build AI agents in
human language, run them across 500+ models without rewriting a line, and set
security, budgets, and access rules once — they apply to every request,
automatically.

This isn't another AI app builder. It's the operating system for an AI-native
company: the place where you describe an agent, hand it to a team, and trust that
every run it makes is routed, optimized, governed, and audited the same way.

## Why DeepMyst

Standing up AI inside a company usually means stitching together a model
provider, a prompt framework, a vector store, a policy layer, an audit trail, and
a billing system — then doing it again for the next team. DeepMyst replaces that
assembly with one platform:

* **Build in human language.** Describe an agent in a sentence. DeepMyst picks the
  right frontend, attaches the right knowledge, and applies your policy.
* **Run on any model.** One OpenAI-compatible API across 500+ models. Swap
  providers with a config flip — your code never changes.
* **Govern once, everywhere.** Identity, policy, security, and audit are inherited
  by every request. Set the rules at the top; they cascade down and can only
  tighten.
* **Scale efficiently.** Every prompt is losslessly compressed and routed to the
  model that does the job best — millions of runs, not millions in spend.

## Built like infrastructure

Every agent runs on the same six-layer stack. A request inherits identity from the
bottom and policy from above, so the stack that runs your first agent runs every
agent.

<CardGroup cols="2">
  <Card title="Shell — how humans use it" icon="terminal">
    Chat, copilot, embeds, and public portals. The surfaces people actually touch.
  </Card>

  <Card title="Agents — first-class processes" icon="boxes-stacked">
    Voice, canvas, document, workflow, kanban, form, and a dozen other shapes.
  </Card>

  <Card title="Memory — what agents know" icon="brain">
    Knowledge graph, documents, datasources, and history every agent reads from.
  </Card>

  <Card title="Skills — what agents can do" icon="wrench">
    Built-in tools, MCP servers, web search, and sub-agents.
  </Card>

  <Card title="Models — what powers their thinking" icon="microchip">
    500+ models behind one OpenAI-compatible API, with automatic failover.
  </Card>

  <Card title="Governance — the rules everything inherits" icon="shield-halved">
    SSO, RBAC, cascading policy, inline KYT security, and audit logged forever.
  </Card>
</CardGroup>

See the full [architecture](/architecture) for how a request flows through the stack.

## What you can build

Every department, every workflow. A single sentence in the builder becomes a
working agent with the right interface for the job.

<CardGroup cols="2">
  <Card title="Engineering" icon="code">
    Bug triage on a severity board, PR reviewers against your style guide,
    structured postmortem intake.
  </Card>

  <Card title="Growth" icon="bullhorn">
    Real-time sales call coaching, brand studios in your voice, pitch decks from a
    one-line brief.
  </Card>

  <Card title="Operations" icon="gears">
    Contract reviewers that cite your procurement policy, onboarding buddies,
    multi-step vendor intake.
  </Card>

  <Card title="Leadership" icon="compass">
    Metrics briefs on warehouse data, working financial models, bespoke strategy
    copilots.
  </Card>
</CardGroup>

Explore the [Agent Builder](/agents) to see the agent types and how they ship.

## Production-grade by default

Agents built in a sentence still behave like production software:

<CardGroup cols="3">
  <Card title="Evaluated" icon="bullseye">
    Every output is scored against your rubric before it reaches anyone.
  </Card>

  <Card title="Governed" icon="shield-check">
    KYT scans every request in under 15ms for injection, PII, and toxicity.
  </Card>

  <Card title="Audited" icon="clock-rotate-left">
    Every run is logged forever and replayable on demand.
  </Card>
</CardGroup>

## Next steps

<CardGroup cols="3">
  <Card title="Agent Builder" icon="wand-magic-sparkles" href="/agents">
    Build agents in human language and ship them to every department.
  </Card>

  <Card title="Governance" icon="shield-halved" href="/governance">
    Set policy, security, and access rules that follow every request.
  </Card>

  <Card title="The OS Stack" icon="layer-group" href="/architecture">
    See how the six layers fit together into one platform.
  </Card>
</CardGroup>