AI Enablement & Bindings Modernisation Proposal

Extension to BDK/LDK Node Bindings Work

---

Executive Summary

What: Two distinct workstreams that together make Bitcoin and Lightning development libraries better for every developer — whether they use AI coding tools or not.

Part 1 — AI Enablement: Make BDK, LDK, and LDK Node discoverable and usable by AI coding assistants (Claude, Cursor, GitHub Copilot, Replit) so developers can build Bitcoin applications faster and with fewer errors.

Part 2 — Bindings Modernisation & Automation: Update UniFFI to the latest approach and replace manual, error-prone release processes with a fully automated CI/CD pipeline across all platforms.

Scope: BDK, LDK, and LDK Node across all platforms (Rust, Android, iOS, React Native, Flutter, Python)

---

---

PART 1 — AI Enablement

---

The Problem

AI Coding Tools Are Now Mainstream

• The majority of developers now use AI coding assistants as a standard part of their workflow
• Tools like Claude Code, Cursor, GitHub Copilot, and Replit AI generate code from natural language
• Developers increasingly say: "If AI can't help me use it, I'll find something else"

Bitcoin/Lightning Libraries Are Invisible to AI

When a developer asks an AI assistant:

"Add Lightning payments to my mobile app"

The AI either:

• Doesn't know about BDK/LDK/LDK Node
• Generates outdated or incorrect code
• Misses critical steps (like calling start() before operations)
• Uses wrong patterns (hardcoded paths, insecure mnemonic storage)

---

The Solution

Make our libraries discoverable and usable by AI coding tools through:

1. Discovery Files (How AI Finds Us)

File	Purpose
llms.txt	AI-readable library summary (like robots.txt for AI)
AGENTS.md	Instructions for coding agents
CLAUDE.md	Claude-specific guidance
.cursor/rules.md	Cursor IDE integration

2. Claude Skills (Direct AI Integration)

Packaged instruction files (.skill) that users upload to Claude.ai. Once uploaded, Claude knows exactly how to use the library correctly.

3. MCP Server (AI Tool Access)

Model Context Protocol server that lets AI agents directly interact with running nodes — create invoices, check balances, open channels through natural language.

4. Structured Documentation

• JSON schemas for all data types
• API references optimised for AI parsing
• Example code with expected outputs

Result: Developer asks "create a Lightning invoice" → AI generates correct, working code.

---

Scope

Libraries

• BDK (Bitcoin Development Kit)
• LDK (Lightning Development Kit)
• LDK Node (Ready-to-go Lightning node)

Platforms

Platform	Bindings	AI Enablement Deliverables
Rust	Core	llms.txt · AGENTS.md · CLAUDE.md · Claude Skill · JSON schemas
Android	Kotlin (UniFFI)	Same + Kotlin-specific examples
iOS	Swift (UniFFI)	Same + Swift-specific examples
React Native	JavaScript/TypeScript	Same + RN-specific examples
Flutter	Dart	Same + Dart-specific examples
Python	UniFFI	Same + Python-specific examples

---

Part 1 Work Items

Per library, per platform:

• llms.txt — AI discovery file
• AGENTS.md — Coding agent instructions
• CLAUDE.md — Claude-specific hints
• .cursor/rules.md — Cursor IDE rules
• library.skill — Claude Skill package
• JSON schemas for all data types
• Platform-specific examples with expected outputs

Quick Wins (Can Start Immediately)

These require no code changes — just documentation that AI tools can read:

1. llms.txt for all libraries
2. AGENTS.md files
3. Claude Skills
4. GitHub metadata optimisation

---

---

PART 2 — Bindings Modernisation & Automation

---

UniFFI Bindings Modernisation

What is UniFFI?

A tool that automatically generates language bindings (Android, iOS, Python, etc.) from Rust code. One Rust implementation → multiple language SDKs.

Current State

• Manual or semi-automated binding generation
• Inconsistent release processes across platforms
• Version mismatches between platforms
• Slow to update when Rust code changes

Proposed Work

• Adopt latest UniFFI patterns and best practices
• Standardise binding generation scripts across all libraries
• Ensure consistent API surface across all platforms
• Document the binding creation process

---

Automated Build & Release Pipeline

Current Pain Points

• Manual binary compilation for each platform
• Release delays due to manual processes (weeks between code change and publish)
• Difficulty keeping all platforms in sync

Proposed Solution

Git Push → CI/CD Pipeline → Automated Builds → Published Packages
                ↓
     ┌─────────────────────────────────┐
     │  • Rust crates (crates.io)      │
     │  • Android (Maven)              │
     │  • iOS (CocoaPods/SPM)          │
     │  • React Native (npm)           │
     │  • Flutter (pub.dev)            │
     │  • Python (PyPI)                │
     └─────────────────────────────────┘

Benefits

• Faster releases (weeks → days, same-day across all platforms)
• All platforms released simultaneously
• Reduced human error
• Easier long-term maintenance

Part 2 Work Items

• GitHub Actions workflows for each platform
• Automated version bumping
• Automated changelog generation
• Release coordination across platforms
• Binary caching for faster builds

---

---

Roadmap

Q2 2026                    Q3 2026                    Q4 2026
───────────────────────────────────────────────────────────────
PART 1                     PARTS 1 & 2                PART 2

AI Enablement Files        UniFFI Modernisation       Build Automation
• llms.txt (all libs)      • Update binding approach  • CI/CD pipelines
• AGENTS.md                • Standardise generation   • Auto-publish
• Claude Skills            • Document process         • Release coordination
• Cursor rules
• JSON schemas             MCP Servers                Content & Outreach
                           • LDK Node MCP             • Tutorials
                           • BDK MCP                  • Workshop updates
                                                      • Community engagement

---

Proposal by Bitcoin Zavior Extension to Spiral BDK/LDK Node Bindings Grant