%20rotate(-45)'/%3e%3ccircle%20class='cls-2'%20cx='1605.42'%20cy='180'%20r='179.96'/%3e%3ccircle%20class='cls-2'%20cx='2051.52'%20cy='2840.24'%20r='179.96'%20transform='translate(-1407.4772%202282.5296)%20rotate(-45)'/%3e%3ccircle%20class='cls-2'%20cx='886.68'%20cy='2862.36'%20r='179.96'/%3e%3ccircle%20class='cls-2'%20cx='2848.41'%20cy='1467.96'%20r='179.96'/%3e%3ccircle%20class='cls-2'%20cx='180'%20cy='2182.96'%20r='179.96'/%3e%3ccircle%20class='cls-1'%20cx='1137.47'%20cy='1926.41'%20r='374.02'/%3e%3cline%20class='cls-1'%20x1='2848.41'%20y1='1467.96'%20x2='1511.59'%20y2='1826.16'/%3e%3cline%20class='cls-1'%20x1='763.36'%20y1='2026.65'%20x2='180'%20y2='2182.96'/%3e%3cline%20class='cls-1'%20x1='1401.99'%20y1='2190.91'%20x2='2051.52'%20y2='2840.24'/%3e%3cline%20class='cls-1'%20x1='872.88'%20y1='1661.97'%20x2='180'%20y2='969.3'/%3e%3cline%20class='cls-1'%20x1='886.68'%20y1='2862.36'%20x2='1040.67'%20y2='2287.68'/%3e%3cline%20class='cls-1'%20x1='1234.28'%20y1='1565.13'%20x2='1605.42'%20y2='180'/%3e%3c/svg%3e)
The Portable Context System for AI.
One portable memory layer for every AI you use. Memplex autonomously builds a private context graph from your AI tools and work systems, routes the right slice to Claude, ChatGPT, Cursor, Codex, Perplexity, and any MCP-compatible AI, and when your chat hits the context wall, the next one opens already knowing where you left off.
Cross-tool by default · Scoped to the task · Auditable end to end · Yours to revoke.
Every AI session starts cold.
You re-explain who you are. What you're building. Who you work with. What's already decided. Memory built into one vendor is trapped in that vendor, and forgotten by every other tool you open tomorrow.
Memplex is the neutral memory layer that travels with you. One context graph, every AI, scoped to the task, governed by you.
Memplex turns scattered work into a living context graph.
Connect once
Authorize the sources you already use, Claude, ChatGPT, Codex, GitHub, Google, Notion, and more. Memplex listens continuously from there. Manual upload is available if necessary.
Memplex builds your graph
People, projects, decisions, services, initiatives, identity, preferences, extracted automatically with paragraph-level provenance. Every entity links back to the source artifact and the moment it was learned.
Your AI gets the right context
When you open Claude, ChatGPT, Cursor, or any MCP-compatible AI, Memplex sends a compact, task-scoped context pack. No identity dump. No cold start. Just the slice that matters.
Your AI never forgets.
Context windows fill up. Sessions end. Models forget. Memplex doesn't, it carries the thread of your work across every session you start, in every tool you use.
Pick up where you left off
Start a new chat in any AI tool, Claude, ChatGPT, Cursor, Codex, Perplexity, or whatever you use next. The decisions you made yesterday, the files you touched, the threads still open are already in context. No re-paste. No re-explain.
Survive the context wall
When a session fills its context window, Memplex carries what matters into the next one. Your work continues across the wall, without you noticing it happen, without losing what was decided.
One memory for every AI
Memplex doesn't care which AI you opened today or which one you'll open tomorrow. The same context feeds all of them. Switch tools whenever, the memory comes with you.
The model's context window didn't get larger. Memplex just made the limit invisible.
A graph that understands relationships, not a pile of embeddings.
Most AI memory tools dump everything into a vector store and hope similarity does the work. Memplex doesn't. We model people, projects, decisions, services, initiatives, and the edges between them, owns, blocks, realizes, caused_by, depends_on, so context is retrieved by traversal, not just resemblance. The result: more accurate answers with smaller, cheaper context packs.
Your context, across every AI you use.
Persistent identity across tools
Your preferences, writing style, projects, and decisions follow you from Claude to ChatGPT to Cursor to Perplexity. One graph, every tool.
Autonomous from day one
Memplex reads from where you already work, your AI tools, your code, your docs, without asking you to upload anything. Authorize once, it keeps up.
Scoped, not dumped
Each AI session gets only the context that fits the task. No identity dump. No oversharing. You see exactly what was sent and why.
Organizational context that respects every permission.
Memplex extends naturally from individuals to teams to entire organizations. Each person keeps a private graph. Teams curate a shared knowledge layer. Admins govern destinations, retention, and audit, all without breaking the source-system permissions you already trust.
Org & team hierarchy
Shared knowledge spaces with approval workflows. Departmental context that stays in the right lane.
Source-truth permissions
Memplex mirrors the ACLs of Google Workspace, GitHub, Microsoft 365, Notion, Linear, and Slack, and revalidates them at every retrieval.
Destination governance
Allowlists for AI tools and models. Per-tool, per-model, per-agent policy. Approved destinations only.
Audit, retention, residency
Full audit log, retention schedules, DLP hooks, legal hold, data residency, SCIM lifecycle, SSO/SAML/OIDC.
Built on open standards. Plugs into the tools you already use.
Memplex speaks the open Model Context Protocol (MCP) plus a clean REST API. Anything that supports MCP, and most modern AI clients now do, connects in minutes.
Built on trust, not black-box memory.
Encrypted vault, envelope encryption at rest
Source-permission aware, Memplex never grants access the source wouldn't
Explicit consent for every source and every destination
Per-tool, per-model, per-agent policy, your AI tools are governed, not free-for-all
Paragraph-level lineage, every memory traces back to its source
One-click export, revoke, and erasure, your context, your call
Prompt-injection neutralization, a poisoned source can't inject hidden instructions into your AI through Memplex. We catch and neutralize them at delivery, before any model sees a thing.
Secrets stay on your machine, API keys, credentials, and connection strings are detected and stripped before any payload leaves your device.
Delete and prove it, permanently erase your data (GDPR / CCPA grade) and receive an auditable receipt as proof.
How Memplex compares.
Most products in the "AI memory" category solve a narrower problem. Here's where Memplex sits next to the three most-asked-about alternatives.
vs. Claude Memory or ChatGPT Memory
Vendor memory works inside one tool. The context Claude remembers about your project doesn't follow you to ChatGPT, and vice versa, by design, because vendor memory is a moat, not a service.
Memplex is the layer underneath every AI tool you use. Authorize a source once; the same context graph feeds Claude, ChatGPT, Cursor, Codex, Perplexity, and any MCP-compatible AI you connect.
vs. Glean and other work-AI suites
Work-AI suites are search engines with a chat UI on top. The chat is their destination, the product, the moat, the lock-in.
Memplex doesn't host a competing chat. Your AI tools stay your AI tools. Memplex provides the governed context plane underneath them, with destination policy and audit per AI tool, model, or agent.
vs. vector RAG and embedding-based memory
Vector search returns documents that are semantically near your query. That's useful for documents and nearly useless for the context an AI needs to act on, decisions, relationships, sequences.
Memplex is graph-shaped, not embedding-shaped. The same substrate Atlassian validated at 150 billion objects: 44% more accurate, 48% fewer tokens than vector RAG. Memplex uses vector search as one of three retrieval channels, never as the whole answer.
Common questions.
What is Memplex?
Memplex is a Portable Context System for AI. It autonomously builds a private context graph from the AI tools and work systems you already use, Claude, ChatGPT, Codex, Cursor, Perplexity, GitHub, Google Workspace, Microsoft 365, Notion, Linear, Slack, and routes the right slice of that context to whichever AI you're currently using, through the open Model Context Protocol (MCP).
How is Memplex different from ChatGPT memory or Claude memory?
Vendor memory is locked to one vendor. Open Claude tomorrow, and ChatGPT forgets everything you told it. Memplex is the neutral memory layer that travels with you across every AI tool. One graph, every model, every client.
How is Memplex different from Glean?
Glean is enterprise search bolted onto corporate documents. Memplex is a living context graph that autonomously listens to your AI tools themselves, and to the work systems behind them, then serves the right context to any AI client via MCP. Different substrate (graph vs. search index), different ingestion model (autonomous vs. indexed), different destination model (any AI, not a single chat product).
How is Memplex different from a vector database or RAG?
A vector database retrieves by similarity. Memplex retrieves by relationships. People, projects, decisions, services, and the edges between them are modeled explicitly, so context arrives smaller, cheaper, and more accurate, with full provenance.
Do I have to upload anything?
No. You authorize the sources you already use; Memplex listens from there.
Which AI tools work with Memplex?
Any tool that speaks the Model Context Protocol (MCP), including Claude (desktop, web, mobile, Claude Code), ChatGPT (web, desktop, mobile), Cursor, Codex, Windsurf, Perplexity, Gemini, Grok, Copilot, and a growing catalog. The same MCP endpoint serves them all.
Is Memplex for individuals or for enterprises?
Both. Individuals get a private context graph that travels across every AI tool they use. Teams and organizations get a governed organizational memory plane with SSO, shared knowledge spaces, policy controls, audit export, and source-permission mirroring.
How does Memplex handle permissions?
Memplex mirrors the access controls of the source systems it connects to, GitHub, Google Workspace, Microsoft 365, Notion, Linear, Slack, and revalidates them at every retrieval. It never grants access the source itself wouldn't. Destination policies are separate and explicit: per-tool, per-model, per-agent.
Is my data used to train models?
No. Memplex does not train foundation models on your data. Ever.
Can I export or delete everything?
Yes. Full Markdown and JSON export at any time. Full erasure (GDPR Article 17 / CCPA) on request, with an auditable receipt.
What happens when my AI chat hits the context window?
Memplex carries your work forward. The next chat you open in the same project arrives already knowing what was decided, which files matter, and what's still open. Works across Claude, ChatGPT, Cursor, Codex, and any MCP-compatible AI tool.
Can I use Memplex from a mobile AI app?
Yes. Mobile AI clients on iOS and Android connect to Memplex through the same protocol every desktop client uses. There's no Memplex mobile app to install; your mobile AI tools authorize once and reach the same context your desktop tools do.
What happens if a connected source revokes my access?
Memplex mirrors your source's permissions and rechecks them continuously. If GitHub revokes you from a repo, or your manager removes you from a Drive folder, Memplex stops serving that context, and the change shows up in your audit log. We don't fake permission control by inventing our own scope tags.
When can I sign up?
Memplex is in private beta now. Use the form below to request early access.
“Memplex is the Portable Context System (PCS) for AI, one portable memory layer for every AI tool a user touches. It autonomously builds a private context graph from a user's AI tools and work systems, then routes the right context slice to any MCP-compatible AI tool: Claude, ChatGPT, Cursor, Codex, Perplexity, Gemini, Grok, without locking users into one vendor's memory.”
Private beta. Small batches.
For individuals and teams building with multiple AI tools. Tell us where context breaks today; that's the bug we're fixing.
No marketing. No bulk rentals. Your data is yours.