Full Pro free for 7 days, no credit card. See pricing →
Features

Five layers of context, every AI can see.

Vilix gives every AI tool you use the same memory, the same project state, the same personal preferences, the same skills, and one inbox to push work to your running agents, automatically, on every reply. Connect once via MCP and the rest happens in the background.

Memory

Cross-tool memory

One memory across OpenClaw, Hermes, Claude Code, Cursor, Codex, ChatGPT, and any MCP tool. Semantic and keyword search over every conversation you have ever had with an AI.

  • Reads the right past context before each AI reply, automatically.
  • Saves the new exchange after each reply, no copy-paste, no buttons.
  • Search by meaning or by keyword, find what you said last month in seconds.
  • Source-aware: Vilix remembers which tool a thought came from, so retrieval stays grounded.
Example
Monday, in ChatGPT

“Brainstorm a launch plan for the new pricing tier.”

Thursday, in Claude

“Continue the launch plan.”

Claude already has Monday's plan, every decision, every constraint, without you pasting a thing.
Projects & Tasks

A project manager every AI can see

A lightweight project manager whose state auto-injects into context. Every connected AI sees what is active, blocked, or done, without you re-explaining the work each time you switch tools.

  • Create projects with project_rules, per-board stack, conventions, tone every AI follows when working on that project.
  • Add tasks with state (todo, doing, blocked, done) and priority (low, normal, high).
  • The active project's full state is injected on each new chat, any AI sees what's open.
  • Full CRUD from any AI tool, your assistant can read, add, update, and close tasks.

Project boards are organized task views any connected AI can read, separate from your always-on memory, which is never limited by how many projects you have. See plan limits on the pricing page.

Example
In Cursor

“Mark the auth task as done and start the billing task.”

Later, in ChatGPT

“What am I working on?”

ChatGPT lists the active project, the current task, and what is still blocked, pulled live, not guessed.
user_rules

Your preferences, applied everywhere

Personal style directives that every AI tool follows. Set them once; they travel with you across every connected AI, on every new chat, automatically. Unlimited and included on every plan, Free, Pro, and Power.

  • Plain-English rules: "Short answers", "No tables", "Always show code first".
  • Applied on every new chat in every connected AI, no re-pasting.
  • Edit, add, or remove from any AI tool at any time.
  • Unlimited on every plan, Free included.
Example
Set once, from any AI

“"Keep answers short. Always call me boss."”

Every chat after

“ChatGPT, Claude, Cursor, Codex, all start with those rules already in context.”

Stop re-pasting your style preferences into every new conversation.
Agent Skills

Teach a procedure once, every AI runs it right

Agent Skills are reusable how-to procedures saved in your Vilix memory. Write down how a recurring task is done, once, and ChatGPT, Claude, Cursor, Codex, GitHub Copilot, and any connected AI loads the exact steps on demand instead of winging it.

  • Save a skill from any AI tool: the steps, the conventions, the gotchas, in plain language.
  • Every connected AI sees your skill index and loads the full procedure exactly when the task comes up.
  • Scope a skill globally, so it is available in every chat, or to a single project.
  • Stop re-explaining the same procedure to every agent. Write it once, it travels with you.
Example
Once, in Claude

“Save a skill: how to check active users. The exact query, what counts as active, who to exclude.”

Weeks later, in Cursor

“Check active users.”

Cursor loads the skill and runs the exact procedure you wrote, the right way, with nothing re-explained.
Agent Inbox

Push work to your own agents from anywhere

Send a message to your own running agents from any connected client, including your phone. Vilix delivers it inside the inbox field of the agent’s next memory read, so an always-on agent picks up the work on its own without you babysitting a session.

  • Send from any connected client (your phone, ChatGPT, Claude, Cursor, Codex), the agent reads it on its next memory check.
  • It rides the next read: the message arrives in the inbox field the next time the agent calls get_context, a search, or get_project.
  • Scope it globally to all your agents, to one project or task, or to a single tool like Claude Code.
  • Normal urgency delivers once, then it is done. Urgent redelivers on every read until the agent calls ack_message or it expires.
Example
From your phone, on the go

“Tell my coding agent: the staging deploy is green, start the smoke tests.”

Minutes later, the agent’s next read

“get_context returns your note in the inbox field, no new session needed.”

Your always-on agent picks the task up on its own next read and runs it, no babysitting and no copy-paste.

One memory. Every AI you already use.

Connect once via MCP. After that, every connected AI shares your memory, your project state, your user_rules, and your Agent Skills, automatically.

Start 7-day free trialTalk to us

No credit card required · Cancel anytime · Export or delete your data whenever you want.