MemoryLake is a multimodal AI memory layer for models and agents. See features, integrations and pricing: free plan, paid plans from $19/month.
Researched and fact-checked by Bartu Cavusoglu · Updated October 26, 2026
MemoryLake is an AI memory infrastructure product. It keeps a persistent memory of your conversations, documents, spreadsheets, audio, video and images, and makes that memory available to whichever AI model or agent you use. The company describes it as neutral by design, so the memory is not tied to one model vendor.
The core idea, according to the company, is that models are general thinking power while memory is what makes a user or organization distinct. Its main consumer-facing product is the Memory Passport, a private, encrypted and portable store of memory. Other products named on the site include WorkBrain, Roster, MemoryLake Router and MemoryLake Compute. The company serves both individuals and enterprises.
MemoryLake ingests data from uploads, conversation history from ChatGPT, Claude and Gemini, cloud drives, databases and APIs. Files are parsed by the company's MemoryLake-D1 engine, which it describes as using visual and logical dual validation for complex Excel and PDF layouts. Parsed content is organized into six typed memories rather than one flat store.
When new memory arrives, the conflict engine compares it with existing memory and flags contradictions. The user chooses how they are resolved, and earlier versions stay in a Git-like history with source attribution. Agents then reach the memory through plugins, MCP, a CLI or a REST API, and every route reads and writes the same memory.
The six memory types are background, fact, event, dialogue, reflection and skill. Fact memory is versioned and traceable to its source, while skill memory holds methods that can be reused across AI tools. Versioning supports branches, rollback and diff views, and audit reports can be exported.
Ingestion covers conversations, text, documents, images, audio and video, plus MySQL, PostgreSQL, Delta Lake, Hudi and Iceberg. Seven public datasets are built in, including academic papers, SEC filings, clinical trials and US patents. On privacy, the company describes triple-party encryption, per-AI access permissions, export and permanent deletion. It reports a 94.03% LoCoMo score and claims of 99.8% recall and 91% fewer tokens, all self-published.
MemoryLake is an AI memory infrastructure product that stores conversations, documents, spreadsheets, audio, video and images as persistent, versioned memory. That memory can be used by any AI model or agent, including ChatGPT, Claude, Gemini and coding agents.
The Free plan is $0 and includes 300,000 tokens per month. Pro is $19 per month for 6.2M tokens, and Premium is $199 per month for 66M tokens. Yearly billing saves about 17%, and extra usage costs $3.125 per 1M tokens through credits.
Yes. The Free plan includes 300,000 tokens per month at no cost and does not require a credit card. Credits are only needed if you exceed the monthly quota.
It works with Claude Code, Codex, Cursor, OpenClaw, Hermes Agent, ChatGPT, Claude, Gemini and Slack. It also offers MCP, a CLI and a REST API, and connects to Google Drive, OneDrive, SharePoint, Dropbox, Lark, DingTalk, WPS 365, MySQL and PostgreSQL.
The company describes triple-party encryption so no single party holds all keys, per-AI access control, one-click export and permanent deletion. It lists ISO/IEC 27001, SOC 2 Type 2, GDPR and PCI DSS, with reports available on request.
Conflicts are detected when a memory is ingested. They can be resolved by recency, source priority, confidence-weighted or manual rules, and both versions stay in the version history with source attribution.
It handles conversations, text, documents, spreadsheets, images, audio and video. Pricing lists PDF, Word, PPT, Excel, CSV, Txt and Markdown processing, and the company's MemoryLake-D1 engine parses complex Excel and PDF layouts.
It suits individuals and teams who use several AI tools and want one shared, auditable memory. The company also targets enterprises, with pages for finance and use cases in legal, research, healthcare and operations.