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

# Welcome to Sparkle

> Sparkle is the control plane for the AI-native SDLC. It makes AI-generated code safer by giving coding agents the context of your repositories before they write code, and by verifying the result after.

Sparkle adds coding guardrails to AI-assisted development and agents so generated code follows your security, architecture, best practices and compliance expectations.

You install Sparkle once — as an IDE extension or through the CLI — and it works inside the tools your team already uses: Cursor, Windsurf, VS Code, Claude Code, Codex, and more.

## Why Sparkle

AI coding tools can generate software faster, but they do not automatically understand how your organization/team expects software to be built.

They often lack the repository context, architecture decisions, security requirements, and engineering standards that experienced developers rely on. The result is code that may work, but still introduces risk, breaks established patterns, or creates more work during review.

Sparkle gives AI coding tools the same organizational context and guardrails your engineering and security teams expect developers to follow.

With Sparkle, teams automatically:

* Apply repository-specific security and engineering guardrails during code generation
* Reduce repeated guidance across developers, repositories, and coding tools
* Catch guardrail violations during pull request review
* See how AI-assisted development is being governed across the organization

Sparkle helps teams move faster with AI without giving up control over how software is designed, generated, and reviewed.

## How Sparkle works

Sparkle applies guardrails across the full lifecycle of an AI coding task:

<Steps>
  <Step title="Context before generation">
    Sparkle builds context from your repositories, architecture, policies, and engineering standards so its guardrails reflect how your codebase is expected to work.
  </Step>

  <Step title="Guardrails during generation">
    When an AI coding task begins: Sparkle understands the user intent, provides prompt & repository-specific guardrails that guide the agent while it plans, generates, and modifies code.
  </Step>

  <Step title="Verification in pull requests">
    Sparkle reviews code changes against the applicable guardrails during pull request checks and surfaces deviations before merge.
  </Step>

  <Step title="Visibility and governance">
    Teams can see which guardrails were applied, where risks were identified, and how AI-assisted development is being governed across repositories.
  </Step>
</Steps>

## Where Sparkle runs

<CardGroup cols={2}>
  <Card title="CLI" icon="terminal" href="/cli/usage">
    Configure terminal-first agents like Claude Code and Codex, or manage Sparkle from scripts and repeatable setup.
  </Card>

  <Card title="IDE extension" icon="plug" href="/extension/usage">
    Sparkle Guardrails for Cursor, Windsurf, and VS Code. Guardrails apply on every AI prompt inside your editor.
  </Card>
</CardGroup>

The CLI and IDE extension connect to the same Sparkle workspace and apply the same guardrails.

## Before you begin

You need:

* A Sparkle account — [sign up for free](https://app.thesparkle.ai)
* A supported IDE or coding agent
* A repository with a git remote (Sparkle resolves guardrails per repo)

## Get started

<Card title="Quickstart" icon="rocket" href="/get-started/quickstart">
  Install Sparkle and apply guardrails to your first AI coding task in a few minutes.
</Card>