---
title: Build with LLMs
description: Use Thru docs, skills, and Explorer MCP effectively when you are
  building with an AI coding agent.
source_url:
  html: https://thru.org/docs/getting-started/build-with-an-llm/
  md: https://thru.org/docs/getting-started/build-with-an-llm.md
---

# Build with LLMs

> **Note:**
>
> This page is a routing and guardrails guide. Use it to pick the right next page, then switch to the linked SDK, ABI, CLI, or MCP references for exact syntax and API details.

## Install Thru Skills

Thru-specific skills live in the public [`Unto-Labs/ai`](https://github.com/Unto-Labs/ai) repository.

Start by installing the `thru-best-practices` skill. It is the highest-value default skill for agentic Thru development and should be part of the agent setup before working on programs, ABIs, SDK integrations, or CLI workflows.

To install the Thru skills repo:

```bash
npx skills add Unto-Labs/ai
```

## Pick Your Path

| If the task is… | Start here | Then use |
| - | - | - |
| Build a first C program | [Quickstart: Build a C Program](https://thru.org/docs/program-development/building-a-c-program.md) | [C SDK](https://thru.org/docs/sdks/c.md), [Recommended Development Pattern](https://thru.org/docs/program-development/program-development-lifecycle.md) |
| Work on CPI or authorization flow | [Cross Program Invocation](https://thru.org/docs/program-development/cross-program-invocation.md) | [C SDK Cross-Program Invocation](https://thru.org/docs/sdks/c-reference/cross-program-invocation.md), [Common Gotchas](https://thru.org/docs/sdks/c-reference/common-gotchas.md) |
| Integrate the hosted wallet in a web app | [Wallet Overview](https://thru.org/docs/wallet/overview.md) | [Embedded Wallet Integration](https://thru.org/docs/wallet/embedded-wallet-integration.md), [Approval and Signing](https://thru.org/docs/wallet/approval-and-signing.md) |
| Write or update an ABI | [Overview](https://thru.org/docs/abi/overview.md) | [Authoring Guide](https://thru.org/docs/abi/authoring-guide.md), [Examples](https://thru.org/docs/abi/examples.md), [Specification](https://thru.org/docs/abi/specification.md) |
| Prove an ABI is correct before deploy | [Validation and roundtrip testing](https://thru.org/docs/abi/validation-and-roundtrip-testing.md) | [ABI Analyze](https://thru.org/docs/cli-reference/abi-analyze.md), [ABI Codegen](https://thru.org/docs/cli-reference/abi-codegen.md), [ABI Reflect](https://thru.org/docs/cli-reference/abi-reflect.md) |
| Make explorer reflection work for a published ABI | [Explorer Compatibility](https://thru.org/docs/abi/explorer-compatibility.md) | [Publishing and Iteration](https://thru.org/docs/abi/deployment-and-program-testing.md), [Explorer MCP Overview](https://thru.org/docs/api-ref/explorer-mcp/overview.md) |
| Publish or upgrade program + ABI | [Recommended Development Pattern](https://thru.org/docs/program-development/program-development-lifecycle.md) | [Publishing and Iteration](https://thru.org/docs/abi/deployment-and-program-testing.md), [Program](https://thru.org/docs/cli-reference/program-commands.md), [ABI](https://thru.org/docs/cli-reference/abi-commands.md) |
| Fund an account so it can pay for test or beta-phase writes | [Faucet](https://thru.org/docs/cli-reference/faucet-commands.md) | [Network](https://thru.org/docs/cli-reference/network-commands.md), [Transaction](https://thru.org/docs/cli-reference/transaction-commands.md) |
| Build a web app or backend integration | [SDK Packages](https://thru.org/docs/program-development/developer-resources.md) | [Web Overview](https://thru.org/docs/sdks/web.md), [Rust](https://thru.org/docs/sdks/rust.md), the package reference page you actually need |
| Debug deployed behavior | [Explorer MCP Overview](https://thru.org/docs/api-ref/explorer-mcp/overview.md) | [Tools Reference](https://thru.org/docs/api-ref/explorer-mcp/tools-reference.md), [Transaction](https://thru.org/docs/cli-reference/transaction-commands.md), [runtime errors](https://thru.org/docs/spec/runtime/errors.md) |

## Use the Docs Site as Context

When an agent is already on a docs page, prefer the docs site’s built-in AI-friendly actions:

- use `Copy for LLM` when you want to paste the full page into a chat
- use `View as Markdown` when you want a cleaner source for an agent to read
- keep the copied context narrow and task-specific instead of pasting large parts of the site

That works best when you start from the smallest page that fully covers the current task.

## Use Explorer MCP for Live Chain Context

For deployed-chain tasks, Explorer MCP is usually the safest way to keep an agent grounded in current state.

Use it for:

- account lookups
- transaction inspection
- recent blocks and recent transactions
- program ABI lookup
- ambiguous identifier search

Start with:

- [Explorer MCP Overview](https://thru.org/docs/api-ref/explorer-mcp/overview.md)
- [Explorer MCP Tools Reference](https://thru.org/docs/api-ref/explorer-mcp/tools-reference.md)
