$ guides / ai-development

Developing with AI

Use AI coding agents to integrate AuthForge licensing into your project faster.

Guide

AuthForge publishes a single, machine-readable documentation bundle at https://docs.authforge.cc/llms-full.txt. It contains the full docs site — SDK reference pages, public auth error codes, best practices, and integration guides — in one file. Point your AI agent at this URL (or paste the prompt below) and it can load everything it needs without you attaching files from GitHub.

For a shorter index, see llms.txt. Cursor and other MCP-capable tools can also use the docs MCP server at https://docs.authforge.cc/mcp to search and fetch pages on demand.

Integration prompt

Paste this into your AI tool with your project open. The agent should fetch the docs itself in Step 1 — you do not need to attach AGENTS.md or other reference files manually.

You are going to integrate AuthForge license validation into my project.
Do not write any code yet. Follow this protocol exactly.

STEP 1 - LOAD REFERENCES
Fetch https://docs.authforge.cc/llms-full.txt and read it end-to-end.
This file is authoritative. If anything below conflicts with it, follow
the docs, not your prior knowledge.

STEP 2 - LEARN MY PROJECT
Look at the project I have open. Identify:
  - Language and framework (e.g. .NET 8 WPF, Python CLI, Electron, Go service)
  - Where startup happens (main, App.xaml.cs, __main__, etc.)
  - How the app currently handles configuration and secrets
  - Whether there is already a user-facing settings / login UI

STEP 3 - IMPLEMENT
For every file you touch:
  - Match my project's existing style (indentation, naming, error handling
    patterns - look at neighbouring files)
  - Use the official AuthForge SDK for my language. If no official SDK
    exists for my stack, say so and stop - do not roll your own client unless explicitly requested.
  - Verify the Ed25519 signature on every response. The SDK does this
    automatically; if you are writing raw HTTP for some reason, do not
    skip it.
  - Handle every error code listed in the docs' "Public auth error codes"
    section, not just the happy path.

After the agent completes Step 2, provide your App ID and App Secret through your project’s existing config mechanism (environment variables, user secrets, etc.) — never commit credentials to source control.

Supported AI tools

This workflow works with Cursor, GitHub Copilot, Claude Code, Windsurf, and any assistant that can fetch URLs or read attached context files.

If your tool cannot fetch URLs at runtime, download llms-full.txt once and attach it to the session instead of relying on Step 1’s fetch.

Tips for better results

  • Open the right project before pasting the prompt so Step 2 can inspect your actual entry point, config, and UI patterns.
  • Name your stack explicitly in a follow-up if it is ambiguous (for example ”.NET 8 WPF” or “Electron + React”).
  • Ask for a plan first — the prompt already defers implementation until after Steps 1–2; keep that order if you iterate.
  • Review error handling against the “Public auth errors” section in llms-full.txt before shipping; generated code often skips hwid_mismatch, session_expired, and rate-limit paths.
  • Each official SDK repo also ships an AGENTS.md with copy-paste patterns and a “Do NOT” checklist — useful as a second pass after the main integration.

Official SDKs

llms-full.txt includes per-language SDK pages. Repositories: