Can I share project rules between different AI assistants?

Yes. You can store project rules in Vilix AI and share them with every supported tool that you connect. Each client needs the Vilix AI connection and custom instructions once. Then you can update the shared rules in your account without maintaining a separate copy in every assistant.

The problem: every AI assistant is an island

Every serious AI tool has its own custom instructions or system prompt field. ChatGPT has “Custom Instructions.” Claude has Projects with custom instructions. Cursor supports project rules. Codex supports project instructions like AGENTS.md. So, when you want every one of these tools to know that your project uses TypeScript strict mode, prefers short answers and has an API backend at a specific base URL, you have to copy the same block of text into every tool and keep them all in sync.

This is not ideal, especially when you're constantly starting new projects or adding new tools to your stack.

How Vilix AI solves it with user_rules and project_rules

Vilix AI has two types of rules, which are stored server-side in your account and returned to connected tools in context requests:

user_rules: personal style directives

These are short personal directives that apply in all projects, everywhere. For example: ‘keep answers under three paragraphs’, ‘never suggest the installation of a new dependency without asking first’, or “use British English spelling.”

You define them once, and connected tools get them when requesting context.

project_rules: per-project stack and conventions

These are attached to a specific Vilix AI project and contain its technical context: the stack, conventions, things the AI should or should not do in this project's code. Here is an example of what project_rules might contain:

This project is a Next.js 14 app with TypeScript strict mode. Always use the App Router. Use named exports for reusable components; keep the default exports required by Next.js route files. Tailwind only, no inline styles.

When you identify the project in a connected tool and request context, the AI can read its rules before replying. Opening a project in the dashboard does not by itself select that project in every client.

How the auto-injection works

Vilix AI is an MCP-native memory layer. Connect and authenticate each supported client at api.vilix.ai/mcp, then add Vilix AI's custom instructions or wire the calls into your agent loop. The sequence uses two tools:

get_context : runs before the AI replies. It gives the AI your recent conversation history, saved memories, semantically related past exchanges, user_rules, and project rules and state relevant to the work you identify.

save_turn: runs after the AI composes its reply, but before it sends that reply. It persists the exact exchange, and the AI then returns the saved text and reuses the chat_id within the conversation.

The custom instructions or your agent's orchestration define this sequence. Verify both calls in a real conversation so that you know that the client is retrieving rules and saving exchanges.

A concrete example: sharing rules between ChatGPT and Claude

Let's say I'm working on a SaaS product. In Vilix AI, I create a project called "Billing service" and add this project_rules:

Stack: Node.js, Stripe, PostgreSQL

Always wrap Stripe calls in try/catch and log the error to Sentry

API routes follow REST conventions, no GraphQL

Short answers preferred, include code snippets when the question is technical

I also have a user_rule: 'Never start a reply with the word "Certainly."'

Now I ask ChatGPT to write a new webhook handler, it reads the project and user rules via Vilix AI before replying. Then I switch to Claude to review the same webhook handler, and Claude reads the same rules from the same Vilix AI account. I get consistent, convention-aware output from both tools without having to repeat myself.

What tools are supported

Vilix AI rules are available in supported clients with a compatible remote MCP connection. These include ChatGPT, Claude, Claude Code, Cursor, Codex, Grok, Manus, GitHub Copilot, Windsurf, Lovable, OpenClaw, and Hermes. After configuration, the client's get_context calls can retrieve your rules.

For a more detailed look at what cross-tool memory enables beyond just rules, see the cross-AI memory overview.

Setting up your first rules

After connecting Vilix AI to your first tool, there are a couple of ways to add rules. You can ask the AI to add a rule and verify that the write succeeded. For example, you could say, "Add a user rule: always prefer async/await over promise chains." The rules workflow is described on the Vilix AI features page.

Rules are stored in your Vilix AI account, and updates become available on subsequent context reads, so you don't need to edit a separate copy in each connected tool.

If you want to have the same conventions in every tool without copy-pasting, the easiest way to start is by trying Vilix AI free and adding your first project_rules in the first session. Configured clients can then retrieve those shared rules when they request context.

Frequently asked questions

Can I share project rules between ChatGPT and Claude?

Yes. Connect both tools to Vilix AI at api.vilix.ai/mcp, authenticate with the same account, and add the custom instructions in each client. Their context requests can then retrieve the shared project rules.

What is the difference between user_rules and project_rules in Vilix AI?

user_rules are personal style directives that apply to every project and every tool. For example, your preferred answer length, tone, or habits you want the AI to avoid. project_rules are attached to a specific Vilix AI project and contain its technical context: the stack, conventions, the dos and don'ts for that particular code.

Do I need to mention my rules in every chat message?

You don't need to repeat the rules in every message, once the client follows the Vilix AI instructions. Its get_context call returns available rules before the reply. Identify the relevant project when necessary, and check the output because retrieving a rule doesn't guarantee compliance.

What happens if I update a rule? Do all tools pick it up?

Yes! Rules live server-side in your Vilix AI account. So, the next time any of your tools call get_context, they will receive the new version of the rules. There is no per-tool update needed.

Which AI tools support Vilix AI rules?

Supported clients include ChatGPT, Claude, Claude Code, Cursor, Codex, Grok, Manus, GitHub Copilot, Windsurf, Lovable, OpenClaw, and Hermes. Use the setup instructions for your client and verify that get_context returns the expected rules.