Persistent Context · AI Memory Assistant

An AI that remembers your work.

Human work isn't a collection of isolated prompts—it is an evolving timeline of calls, documents, code commits, and architectural trade-offs. Maple is an AI assistant with persistent memory that indexes your workspace locally, giving you an AI knowledge assistant that recalls past context on demand.

Get early access How AI memory works

Indexed on your hardware · SQLite vector embeddings · Private & offline-ready

Maple AI Memory Assistant doodle illustration AI MEMORY

The amnesia problem in modern AI

Standard chatbots restart at zero every time you close the tab. Here is why persistent memory changes how you work.

Multi-Day Recall

Ask: "What were the three objections our client raised during Tuesday's product sync?" Maple retrieves exact quotes from your meeting transcripts and synced docs without manual note-taking.

Connected Knowledge Graph

Maple connects decisions across disparate tools. A mention in Slack connects to a pull request in GitHub and a Figma specification reviewed yesterday.

Zero Prompt Engineering

You never need to start prompts with "Act as an expert in project X, which does Y...". Maple already knows the names, specs, and history of your active projects.

How Maple's AI memory operates under the hood

Engineered in native Rust with local vector embeddings stored in encrypted SQLite files on your SSD.

1. On-Device Ingestion

Maple captures meeting audio via loopback drivers, indexes selected local project directories, and observes active screen window titles locally.

2. Local Vector Embedding

Text is quantized and embedded on your hardware using Apple Neural Engine (macOS) or Tensor cores (Windows), never transmitted to a server for indexing.

3. Sub-100ms Hybrid Search

Combining BM25 full-text indexing with vector similarity ensures instant recall for specific keywords, dates, names, and concepts.

4. Ephemeral Context Delivery

Only the precise context chunk needed to answer your prompt is sent to the AI inference model, with zero data retention or training.

Complete workspace intelligence

Memory is just one facet of Maple's unified desktop architecture.

Give your AI the memory it has always needed.

Join our early access program for macOS and Windows.

Get early access