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

# What is Svantic

> Not a chatbot framework — a control plane for autonomous work. Build agents from existing services, orchestrate them, govern every decision.

# What is Svantic

**From pilot to production. That's the hard part. That's what we built.**

Every company that scales automation hits the same wall. Standard software works perfectly — until something unexpected happens. A vendor changes a portal, a regulation shifts, a data format evolves, and the system breaks. A human steps in, logs into three systems, cross-references data, and patches the flow. Repeat that across hundreds of services, each changing on their own schedule, and you get a maintenance burden that grows linearly with coverage.

The space between deterministic fragility and probabilistic unreliability is where you deploy resources building workarounds, designing fallback processes, and duct-taping systems together.

#### Svantic closes that gap.

## What it does

Svantic turns your existing services into AI agents — with the orchestration, observability, and governance to run them for real. Your CRMs, ticketing systems, monitoring tools, and internal APIs become intelligent agents that communicate, plan, and act. You ship agents instead of spending months on plumbing.

Three things make this work:

<Columns cols={3}>
  <Card title="Tools Execute Locally" icon="shield">
    Every other agent framework assumes tools run on the server. Svantic flips this. Your agent owns its tools locally — the mesh never touches your credentials, filesystem, or database. This is what makes agents usable in regulated industries where data sovereignty matters.
  </Card>

  <Card title="Five Lines to Agent" icon="zap">
    Define capabilities, register with the mesh, and your existing service is an AI agent — discoverable, orchestratable, governed. A developer adds a new capability by writing a handler. A DevOps engineer changes what an agent can do by editing a config and restarting. The barrier is near zero.
  </Card>

  <Card title="Knowledge Compounds" icon="brain">
    Most AI starts from zero every session. Svantic maintains a persistent, vector-indexed knowledge store. Every execution teaches the system — what worked, what failed, what to avoid. The practical consequence: the cost of the next task decreases over time, not increases. This isn't a feature. It's the core moat of the platform.
  </Card>
</Columns>

## How you use it

<Columns cols={2}>
  <Card title="Build Agents" icon="code">
    Install `@svantic/sdk`, define capabilities as typed functions, and call `agent.start()`. Your service is now an A2A-compliant agent on the mesh. Or skip writing code entirely — **Forge** generates agent code from an OpenAPI spec, TypeScript source, or a plain English description in seconds.
  </Card>

  <Card title="Manage Knowledge" icon="database">
    Upload documents, playbooks, and domain context into the **knowledge store**. Agents query it before acting — retrieval-augmented generation with no extra infrastructure. The system learns from every execution, detects stale knowledge, and compounds what works. You control what goes in, what gets updated, and what gets retired.
  </Card>

  <Card title="Orchestrate" icon="network">
    Every agent, session, tool call, and knowledge base is accessible through one **REST API** and one SDK. Start a session, send a message, and the platform figures out which agents to call, in what order, with what context. You don't write routing logic. You don't manage agent-to-agent communication. The mesh handles it.
  </Card>

  <Card title="Your Infrastructure" icon="server">
    Agents run wherever you run them — your cloud, your VPC, your laptop. Svantic coordinates them, but your code and credentials never leave your environment. Knowledge and session data live on the platform, encrypted and isolated per organization. The platform sees capabilities and context — not your raw data.
  </Card>

  <Card title="Open Standards" icon="globe">
    Agents communicate via the **A2A protocol** — Google's agent-to-agent standard. Any compliant agent can join the mesh, whether built with Svantic or not. Swap LLMs, swap providers, swap infrastructure. The platform adapts.
  </Card>
</Columns>

**[Get started →](getting-started)**