LLM Integration

This section explains our recommended approaches for using LLMs and AI-assisted development tools within the Hathor Network ecosystem.

This article covers two complementary workflows:

1. Hathor Skills, which are code skills that can help LLMs understand Hathor-specific development rules and workflows.
2. Hathor Forge MCP integration, which lets an AI assistant control a local Hathor development environment through MCP tools.

Integration options

Integration	Use it when	Article
hathor-wallet-headless skill	You want help building wallet automation, payment backends, token flows, NFTs, atomic swaps, multisig, or nano contract calls through Headless Wallet.	Headless Wallet Skill
hathor-blueprint skill	You want help creating, reviewing, or debugging Hathor nano contract blueprints written in Python 3.11.	Blueprint Skill
Hathor Forge MCP server	You want an AI assistant to control a local Hathor development environment, create wallets, fund them, send local transactions, inspect blocks, or test APIs locally.	Hathor Forge LLM Integration

What are Hathor Skills?

Hathor Skills are reusable code skills designed to help developers work with the Hathor Network ecosystem more effectively while leveraging different language models such as Claude, GPT, Deepseek, etc.

Each skill is a structured instruction package with a SKILL.md file, which defines the skill name, when it should activate, domain-specific rules, examples, and supporting references.

The hathor-skills repository currently includes:

Skill	Purpose
hathor-wallet-headless	Helps the LLM work with the Hathor wallet-headless HTTP API for wallets, balances, transactions, tokens, NFTs, nano contracts, atomic swaps, multisig, and related flows.
hathor-blueprint	Helps the LLM create, review, and debug Hathor nano contract blueprints written in Python 3.11.

The LLM can automatically load skills from a personal skills directory or a project-level skills directory. You can also invoke a skill explicitly by name.

Skill Integration

Before using Hathor's skills in any LLM workflow, you must decide where the skill should live.

• You can use a personal installation when you want the skill available in all your Claude Code sessions.
• We also recommend a project-level installation when you want the skill available to everyone working in a specific repository.
• Additionally, you can use a manual context integration when your LLM does not support Claude Code skills directly.

warning

Installing the LLM skill does not install or run the Headless Wallet itself. The skill helps your LLM work with service, but the service itself must still be installed, configured, and ran separately.

Repository Cloning

Start by cloning the repository:

git clone https://github.com/HathorNetwork/hathor-skills.git
cd hathor-skills

Personal installation

Install the skills into your personal Claude Code skills directory:

mkdir -p ~/.claude/skills

cp -r skills/hathor-blueprints ~/.claude/skills/hathor-blueprint
cp -r skills/hathor-wallet-headless ~/.claude/skills/hathor-wallet-headless

The installed folder name should match the name field inside each SKILL.md file.

Source folder	Installed folder
skills/hathor-blueprints	~/.claude/skills/hathor-blueprint
skills/hathor-wallet-headless	~/.claude/skills/hathor-wallet-headless

Restart Claude Code after installing the skills.

Project-level Installation

To make the skills available to everyone working in a repository, copy them into the project-level .claude/skills directory:

mkdir -p .claude/skills

cp -r /path/to/hathor-skills/skills/hathor-blueprints .claude/skills/hathor-blueprint
cp -r /path/to/hathor-skills/skills/hathor-wallet-headless .claude/skills/hathor-wallet-headless

git add .claude/skills
git commit -m "Add Hathor Claude Code skills"

Restart Claude Code from inside the project repository.

This is useful when the team should share the same LLM behavior for Hathor development.

Local Development with Links

If you are actively editing the skills or want to avoid copying files, link the skills instead:

mkdir -p .claude/skills

ln -s /path/to/hathor-skills/skills/hathor-blueprints .claude/skills/hathor-blueprint
ln -s /path/to/hathor-skills/skills/hathor-wallet-headless .claude/skills/hathor-wallet-headless

Restart Claude Code from inside the project repository.

Skill Activation

Claude Code can activate a skill automatically when your request matches the skill description.

User request	Likely skill
“Create a token with wallet-headless.”	hathor-wallet-headless
“Help me call /wallet/simple-send-tx.”	hathor-wallet-headless
“Design a backend payment flow using Headless Wallet.”	hathor-wallet-headless
“Write a Hathor nano contract blueprint.”	hathor-blueprint
“Why does my blueprint fail with forbidden syntax?”	hathor-blueprint
“Review this nano contract for persistent storage issues.”	hathor-blueprint

You can also invoke a skill explicitly:

/hathor-wallet-headless

/hathor-blueprint

Use skills with other LLMs

Claude Code can load skills automatically. Other LLM tools may not support Hathor skills directly.

For other LLMs, use the skill as structured context:

1. Open the relevant skill from the hathor-skills repository.
2. Attach or paste the skill instructions into your LLM session.
3. Tell the LLM to use the skill as its main Hathor-specific development guidance.
4. Ask the LLM to reference the official Hathor documentation when endpoint schemas, setup steps, or protocol behavior must be verified.

Example prompt:

I am building with Hathor Network.

Use the attached Hathor skill as your main development guidance.

When helping me:
- focus on architecture, implementation workflow, debugging, and code review
- do not invent endpoint schemas or protocol behavior
- tell me when I need to verify details in the official Hathor documentation
- explain assumptions clearly
- avoid exposing secrets, seeds, private keys, or API keys

Choosing the right workflow

Objective	Recommended workflow
Build a backend wallet integration	Use the hathor-wallet-headless skill.
Create a payment endpoint	Use the hathor-wallet-headless skill and verify endpoint details in the Headless Wallet API reference.
Create or debug a nano contract blueprint	Use the hathor-blueprint skill.
Validate blueprint rules before deployment	Use the hathor-blueprint skill and official nano contract documentation.
Run a local Hathor stack	Use Hathor Forge.
Let an AI assistant create wallets and send local transactions	Use Hathor Forge MCP.
Test wallet-headless locally without using public networks	Use Hathor Forge and the Headless Wallet skill together.

Recommended development flow

A practical AI-assisted Hathor workflow usually looks like this:

1. Choose the component. Decide whether you are working with Headless Wallet, blueprints, or a local Forge environment.
2. Open the right article. Use the skill-specific article or the Forge MCP article.
3. Ask for architecture first. Do not start with code generation.
4. Ask for an implementation plan. Include files, configuration, validation, testing, and security checks.
5. Generate code in small pieces. Ask the LLM to create one module, service, or contract at a time.
6. Review the output. Ask the LLM to check its own assumptions against the relevant Hathor documentation.
7. Test locally. Use testnet or Hathor Forge before using production infrastructure.
8. Audit secrets and access. Do not expose seeds, private keys, API keys, wallet services, or MCP servers.

Starter Projects

Use these starter projects to experiment with the LLM integration workflows locally.

Starter project	Description
Blueprint Token Vault Starter	Minimal Python blueprint starter for a token vault use case.
Headless Wallet Payment Starter	TypeScript backend starter for a Headless Wallet payment endpoint.
Hathor Forge MCP Local Flow Starter	TypeScript local wallet flow starter for Hathor Forge MCP.
All LLM Starter Projects	Combined archive with all starter projects.

Next steps

Start with the article that matches your task:

• Use Headless Wallet Skill to build wallet automation, payment backends, token flows, NFTs, atomic swaps, multisig flows, or wallet-driven nano contract calls.
• Use Blueprint Skill to create, review, or debug Hathor nano contract blueprints.
• Use Hathor Forge LLM Integration to let an AI assistant control a local Hathor development environment through MCP.