01 / Work wherever you want
Use the AI tool that makes sense for the task. Brainstorm in ChatGPT. Write and structure in Claude. Explore marketing ideas in Grok. Execute workflows with Manus. You don't need to change how you work.
ONE PROJECT.
ONE MEMORY.
EVERY AI.
Memrix gives your project one persistent memory layer across the AI tools you use, so you can move between them without rebuilding the context every time.
AI can propose what should be remembered. You decide what stays.
01 / The Problem
Your project has one context.
Your AI tools don't.
ChatGPT
"Here is the marketing plan for the launch..."
Context A (Isolated)
Claude
"Based on the product requirements document..."
Context B (Isolated)
Grok
"Analyzing competitive landscape..."
Context C (Isolated)
Decision drift
An AI makes a reasonable recommendation based on incomplete context. It conflicts with a decision you already made somewhere else. The outputs look fine, but the project starts becoming inconsistent.
Memrix is a persistent project memory layer.
Instead of each AI tool maintaining an isolated understanding of your project, Memrix provides a shared memory layer that can be used across your entire workflow.
ChatGPTThink
ClaudeWrite
GrokExplore
ManusExecute
02 / The Insight
The memory should belong to the project, not the AI platform.
ChatGPT
Claude
Grok
Manus
MemrixPersistent Project Memory
03 / Workflow
From fragmented AI sessions to shared project memory.
02 / Important project context becomes memory
As your project develops, important information can become part of its persistent memory. That can include: Decisions, Goals, Constraints, Product context, Strategy, Preferences, Discoveries, and Important project knowledge.
03 / Memrix becomes the shared memory layer
Instead of rebuilding your project's context for every AI platform, Memrix provides a persistent project memory that can be carried across your AI workflows.
04 / Switch AI tools without starting over
Move from one AI platform to another while keeping the project's important context available. The task changes. The AI changes. The project memory remains.
05 / You remain in control
Memrix V1 is intentionally human-controlled. The system does not silently decide what should become permanent project memory. AI proposes → Human approves → Memrix remembers.
A Real AI Workflow
Without Memrix
ChatGPTBrainstorm ideas
↓ loses context
ClaudeWrite PRDs & decisions
↓ loses context
Grok / ManusMarketing & execution
With Memrix
ChatGPTBrainstorm
Memory Created
Project Goal: Optimize landing page conversion rate.
ClaudeWrite PRD
Memory Created
Decision: Use Tailwind CSS tokens only.
Grok / ManusExecution
04 / State
What Memrix remembers.
The goal isn't to remember everything. It's to preserve what would otherwise need to be explained again.
Context
Project Context
What the project is, what it is trying to accomplish, and the context surrounding it.
Decision
Decisions
Important decisions that have already been made and, where useful, the reasoning behind them.
Goal
Goals
What the project is working toward and what outcomes matter.
Constraint
Constraints
Requirements, limitations, rules, preferences, and boundaries that should continue to influence future work.
Context
Product & Strategy Knowledge
Important information about the product, market, users, positioning, or strategy.
Context
Discoveries
Useful findings, insights, research, and knowledge uncovered while working on the project.
Context
Ongoing Context
Important information that needs to carry forward into future AI-assisted work.
Stop being the project's memory layer.
Context-loss loop
A new AI tool doesn't have the context from previous work. → You reconstruct it. → Some context is missing. → The AI starts from an incomplete understanding.
Re-explanation loop
You repeatedly explain: "Here's what we're building." "Here's what we decided." "Don't change this." "Remember this constraint." The work gets slower as the project grows.
Copy-paste loop
Information moves manually between: AI → notes → AI → document → AI → AI. The workflow becomes increasingly dependent on human handoffs.
Decision-drift loop
A new AI receives incomplete context. → It makes a reasonable recommendation. → The recommendation conflicts with an earlier decision. → The project starts moving in multiple directions.
What is persistent project memory?
Project-level information that remains available across AI-assisted sessions and platforms instead of being limited to a single conversation.
A conversation ends.A session changes.A tool changes.Your project continues.
Distinction
Your AI can remember you without remembering your project.
AI / user memory
Designed primarily around information an AI platform retains about a user or their interactions.
Conversation context
Information available within a particular conversation or working context.
Project memory
Information intentionally associated with a specific project and designed to persist as that project evolves.
Comparisons
Memrix vs. the alternatives.
Memrix is not trying to compete with every tool that stores information. Its focus is narrower: Persistent memory for projects working across multiple AI tools.
ChatGPT Memory
ChatGPT's memory can help ChatGPT retain information about a user and their interactions. Memrix is different in focus: its core purpose is to maintain project-owned memory that can be used across AI platforms, rather than keeping the project context tied primarily to one AI environment.
Claude Context / Compaction
Claude can work with substantial conversation and project context, and context-management mechanisms help maintain usable information during AI work. Memrix addresses a different layer: keeping important project memory persistent beyond an individual AI environment so that the project can move between tools.
Notion / Obsidian
Notion and Obsidian are excellent systems for human-managed notes, documents, and knowledge. Memrix is different in purpose: it is designed specifically around the memory layer required for AI-assisted project continuity.
Manual Markdown / .md / Handoff Files
A project can already maintain its context through files such as context.md. This can work. But maintaining and transferring those files manually creates another workflow for the human to manage. Memrix aims to make project memory a dedicated product layer rather than a repeated manual handoff process.
Similarly Named Products
Memrix is Memrix by Evolv Studio. Its product category is: Persistent Project Memory. Its V1 focus is: One persistent project-memory layer across multiple AI tools.
Built for people who work across AI.
Founders
Use different AI tools for strategy, product, research, writing, and execution without constantly rebuilding the company's context.
Product Builders
Carry product decisions and requirements across research, planning, specification, and development.
Developers
Preserve architectural decisions, technical constraints, requirements, and project context across AI coding workflows.
Marketers
Move between AI tools for research, strategy, copy, content, and execution without repeatedly rebuilding the brand and campaign context.
Researchers
Keep important findings, assumptions, decisions, and evolving knowledge available across AI-assisted research workflows.
AI-Native Teams
Create a shared memory layer for projects increasingly distributed across multiple AI tools.
Why Now
We are entering a multi-AI workflow.
The old question
"Which AI should I use?"
The new question
"Which AI should I use for this task?"
Different models and products have different strengths. You may prefer one for reasoning, another for writing, another for research, and another for execution.
Who remembers the project when the AI changes?
Memrix V1
One memory layer across your AI tools.
V1 is designed around:
- Persistent project memory
- Cross-platform project context
- Important decision preservation
- Goals and constraints
- Project knowledge
- Human approval of proposed memories
The V1 Principle
AI Proposed
Save target audience as: 'B2B SaaS founders scaling past $1M ARR'.
Memrix Remembers.
This human-in-the-loop model gives you control over what becomes persistent memory.
First build the memory layer. Then make it proactive.
V1: Shared Project Memory
Establish a reliable memory layer that allows important project context to persist across AI tools.
One project. One memory.
V2: Proactive Project Memory
The longer-term direction is to make that memory layer increasingly proactive.
- Automatic context surfacing
- Detecting changes and conflicts
- Connecting related decisions
V2 is future direction.
The Origin
A problem people were already trying to solve.
We shared the problem on LinkedIn. The comments revealed that people were already building complicated workarounds to manage context across different AI sessions.
DS
Dheeraj Sachdeva
MarTech & Operations Leader
“Coding agents always consider the codebase as the source of truth. Documentation should not just be passive docs; it should hold the project's live working state.
4712
HC
Hana Carqueja
Data Leader
“State, decisions and testing should all live separately. A single massive context dump doesn't scale. Different memory types need structured separation.
8321
DH
David Ingemar Hedin
Product Development Manager
“Continuously document manually what matters. Builders understand this, but they hate maintaining it manually. Automation should synchronize context silently.
6215
Founder Note
Why I'm building Memrix
I've been using AI across almost every part of the work I build.
The problem is that the work rarely stays inside one AI tool.
I might use ChatGPT to think through an idea, Claude to turn it into a PRD or make a product decision, then switch to another tool for research, marketing or execution.
The project keeps moving.
The context doesn't.
I kept finding myself copying decisions between chats, re-explaining the same project, and checking whether a new AI response still matched what we'd already decided. That's where Memrix started.
The idea is simple: the project should have a memory of its own. The AI tools can change, but the important context should stay with the project.
Memrix is now at the V1 founding tester stage.
I'm looking for people who already work across multiple AI tools to use it on real projects, challenge the assumptions, and help shape what a useful project memory layer should actually remember.
Manav Kain
Founder, Evolv Studio
Answers
Help build the memory layer.
Memrix is still early. We're looking for the first people who regularly work across multiple AI tools and understand how painful fragmented project context can become.
As a founding tester, you'll help answer:
- What project information actually needs to persist?
- How should memory move between AI tools?
- What should AI propose as memory?
- What should humans control?
Stop carrying your project's memory between AI tools.
One project.
One memory.
Every AI.
Memrix
by Evolv Studio