> ## Documentation Index > Fetch the complete documentation index at: https://policykit.xyz/llms.txt > Use this file to discover all available pages before exploring further. # Overview > Understanding PolicyKit's architecture and design principles # Overview PolicyKit is a comprehensive framework for enforcing transaction policies on smart accounts and AI agents. It combines on-chain smart contracts with off-chain computation to provide a flexible, secure, and decentralized policy enforcement layer. ## Design Principles ### Policies as Code Policies are defined programmatically in TypeScript, not configured through a UI. This means you can: * **Version control** your policies alongside your application code * **Test** policies locally before deploying * **Review** policy changes through pull requests * **Compose** policies from reusable building blocks ### Non-Custodial PolicyKit never takes custody of private keys or account access. Account owners register policies on-chain and can update or remove them at any time. The policy enforcement layer operates as a guard or module that validates transactions before execution. ### Decentralized Evaluation Off-chain rules are evaluated by the [Lit Protocol](https://litprotocol.com/) network — a decentralized network of nodes that execute code in secure enclaves. This ensures that off-chain evaluation is: * **Censorship resistant** — no single party can block evaluation * **Tamper-proof** — computation happens in TEEs (Trusted Execution Environments) * **Verifiable** — results are signed by the Lit network and verified on-chain ### Fail-Safe Design PolicyKit supports two fail modes: * **Closed** (default): If the off-chain evaluation layer is unreachable, transactions are blocked. Best for high-security use cases. * **Open**: If unreachable, transactions are allowed with on-chain-only checks. Useful when availability is critical. ## Architecture PolicyKit consists of four main components: ### 1. SDK (`@policy-kit/sdk`) The TypeScript SDK is the primary interface for building and managing policies. It provides: * `PolicyBuilder` — Fluent API for constructing policies * `PolicyKit` — High-level client for deploying and managing policies * `PolicyEngineClient` — Typed wrapper for smart contract interactions * `PolicySimulator` — Local evaluation engine for testing * `LitClient` — Integration with Lit Protocol v8 * `IPFSClient` — IPFS pinning and retrieval ### 2. Smart Contracts (`@policy-kit/contracts`) Solidity contracts deployed on-chain that enforce policy rules. Available as an [npm package](https://www.npmjs.com/package/@policy-kit/contracts) for use in your own Solidity projects: * `PolicyEngine` — Core contract that stores and evaluates policies * `PolicyGuard` — Guard module for smart accounts * `PolicyKit7579Module` — ERC-7579 compatible module * Rule evaluators — Individual contracts for each rule type ### 3. Lit Actions (`@policy-kit/lit-actions`) TypeScript code executed on the Lit Protocol network for off-chain rule evaluation. Available as an [npm package](https://www.npmjs.com/package/@policy-kit/lit-actions) containing the pre-built action bundle and source rules. The Lit Action: 1. Fetches the full policy from IPFS 2. Evaluates Tier 3 (off-chain) rules 3. If all rules pass, threshold-signs an EIP-712 `PolicyApproval` 4. Returns the signed attestation for on-chain verification ### 4. CLI (`@policy-kit/cli`) A command-line tool for common policy operations: * Initialize new policies from templates * Deploy policies (IPFS + on-chain) * Simulate transactions against policies * Inspect and manage on-chain policies ## Supported Chains PolicyKit currently supports: | Chain | Status | | ------------ | --------- | | Base | Supported | | Base Sepolia | Supported | ## Next Steps Install the SDK and CLI. Build your first policy.