No charge today. Cancel anytime.
One memory.
Every AI. $20 a month.
Start in ChatGPT, continue in Claude, and bring the same context wherever you work next.
$20/mo
after a 14-day free trial
No charge today. Cancel anytime.
- ✓Transfer what ChatGPT knows about you to Claude
Copy one prompt into ChatGPT, review what comes back, and move it into your MemoryRouter vault.
- ✓One personal memory across ChatGPT, Claude, Codex, and OpenClaw
Connect the AI you use most today. Add the others whenever you want, same memory every time.
- ✓Review everything before it saves
Nothing writes to your memory until you have seen it and approved it.
- ✓Private, encrypted personal memory
Your vault is encrypted and isolated to your account. It is not shared, trained on, or visible to anyone else.
- ✓Delete your memories anytime
Wipe your entire vault from the dashboard whenever you want. No lock-in, no waiting period.
- ✓Built for normal everyday use
No API keys to manage, no infrastructure to configure. Sign up, connect your AI tools, and go.
Includes 200M memory tokens each month. Additional usage is $0.50 per 1M tokens. Most individual users will never need to think about this.
Need shared memory for your company? Explore Enterprise →
New here? Learn what AI memory is, see what MemoryRouter does, or connect the MCP memory server.
Your memory, your terms
"What happens to what I share?"
The most important question when you give an AI a personal memory. Here's the straight answer.
👁️
You approve what gets remembered
Nothing saves to your memory until you have reviewed and approved it.
🔒
Your memories stay private
Encrypted at rest and in transit, isolated to your account, and never shared or trained on.
🚪
Leave whenever you want
Cancel anytime from your dashboard, and delete your entire vault whenever you decide to.
Pricing Questions
What does MemoryRouter remember?▼
The context you approve: preferences, ongoing projects, people, and background you want your AI to already know. You see and approve it before anything saves.
Which AI tools can I use it with?▼
ChatGPT, Claude, Claude Code, Codex, and OpenClaw all connect to the same personal memory. Connect the one you use most today, then add the others anytime.
Can I import old ChatGPT conversations?▼
Yes. Import exact details from your ChatGPT history using the deterministic importer.
Do I approve memories before saving?▼
Yes. Every write goes through a review step. Nothing lands in your vault until you say yes.
What happens after the free trial?▼
You get 14 days free with no charge today. After that, it is $20 a month unless you cancel first. Cancel anytime before the trial ends and you will not be billed.
Can I cancel and delete my memories?▼
Yes, both, anytime. Cancel from your dashboard, and delete your entire memory vault whenever you want. It is your data.
What are the 200M memory tokens?▼
Every plan includes 200M memory tokens each month, the storage and retrieval that power your memory. Additional usage is $0.50 per 1M tokens. Most individual users will never need to think about this.
Is there a team or company plan?▼
Yes. If you need shared memory for your company, explore Enterprise.
Transfer my memories
14-day free trial. No charge today. Cancel anytime.
Straight answers to the questions people ask
The honest version — what MemoryRouter does, what it does not, and exactly how your memory is handled.
Does this import every word of every ChatGPT conversation?▼
No. MemoryRouter does not scrape or copy your full ChatGPT history. For the ChatGPT-to-Claude transfer, we give you a prompt to paste into ChatGPT; ChatGPT writes a concise summary of what it knows about you, and you review and approve everything before it saves to your vault. Connecting the ChatGPT MCP connector by itself does not import or expose old conversations — connection and historical import are independent.
What exactly does the transfer capture — the "supported exact details"?▼
It captures durable context about you — how you like answers written, what you are working on, and the people, projects, and preferences you have shared — as concise memories you approve, not raw transcripts. In coding CLIs the capture is deliberately narrow: the Claude Code and Codex hooks store one user prompt and one final assistant response per completed turn and exclude tool calls, intermediate steps, commands, patches, compaction, and subagent transcripts. Routine messages and full transcripts are never silently captured.
Is this the same as Claude's native import-memory tool?▼
No. MemoryRouter is a separate, portable memory vault you own — not Claude's built-in memory and not ChatGPT's built-in memory. Both can be enabled at the same time. Connecting MemoryRouter does not read, import, or replace a provider's built-in memory. The point of MemoryRouter is that the same vault works across ChatGPT, Claude, Codex, and OpenClaw, so your memory is not locked to one vendor.
Does ChatGPT or Claude automatically call MemoryRouter every time?▼
On hosted chat surfaces (ChatGPT, Claude, Cowork), no — the model decides when to call a MemoryRouter tool, and writes require your confirmation. There is no lifecycle hook that forces recall or capture on every turn, so ask explicitly when you want a save or a search. Local coding integrations are different: the Claude Code and Codex CLI plugins and the OpenClaw plugin use deterministic lifecycle hooks that run automatically at named events.
Where is my data processed and stored?▼
MemoryRouter receives your request, retrieves relevant memory from your own vault, and stores useful new context in that same vault. Your data is encrypted in transit and at rest, and each vault is isolated to your account — one Memory Key points to exactly one vault, and requests with another key cannot read it. For inference, you can use bring-your-own-key pass-through so your provider key calls the model directly; with the OpenClaw plugin your inference and provider keys stay in OpenClaw and never route through us.
Can I export or delete my memories?▼
Yes. You can remove individual memories in the dashboard at any time, and you can wipe your entire vault — the MCP
forget_all_memories action deletes the whole selected vault and requires the exact confirmation phrase DELETE ALL MEMORIES. MemoryRouter also supports direct memory search and JSONL upload today for export/backfill. Disconnecting a connector does not delete your account or vault.What happens when two memories conflict?▼
Retrieval is semantic and relevance-ranked: MemoryRouter surfaces the highest-signal context for the current request rather than dumping your entire history, so recent and relevant memories carry the most weight. New context is saved as additional approved memories, and because you approve writes and can edit or remove any individual memory in the dashboard, you stay in control when older and newer facts disagree.
Does this work with free ChatGPT and Claude plans?▼
It depends on the host. On Claude, the free plan supports one custom connector, and Pro and Max support custom connectors too. On ChatGPT, the Developer-mode MCP connection currently requires an eligible plan — OpenAI's current guidance lists Plus, Pro, Business, Enterprise, and Education — and managed-workspace admins can still restrict it. The MemoryRouter plan itself is a separate $20/month subscription for your portable vault.
Is MemoryRouter affiliated with React Router, Supermemory, or the AI providers?▼
No. MemoryRouter is an independent product and is not affiliated with React Router, Supermemory, OpenAI, or Anthropic. It connects to ChatGPT, Claude, Codex, and OpenClaw through their supported connector and plugin mechanisms; those company and product names are used only to describe compatibility.
What is included in $20 a month?▼
The Starter plan is $20/month after a 14-day free trial, with no charge today and cancel anytime. It includes 200M memory tokens each month; additional usage is $0.50 per 1M tokens, and there is no markup on inference — you bring your own provider keys and pay providers directly. Shared team memory is available on Enterprise, which is contact-priced.
Answers verified against the live product and docs. Last corrected 2026-08-08.