# Multiplist — Full Documentation for LLMs > Multiplist is a meaning operating system that solves AI amnesia — the systematic loss of insight buried inside conversations, notes, and documents. It extracts decisions, frameworks, and key passages from AI conversations and organizes them into a persistent, searchable vault with full provenance and traceability. ## Product Overview Multiplist exists to solve AI amnesia: the systematic loss of insight buried inside conversations, notes, and documents. Where most AI systems generate output and let the thinking disappear, Multiplist turns conversations into durable, traceable knowledge that can be revisited, recombined, and built on. Our core belief is that intelligence becomes more valuable when it compounds. Multiplist preserves decisions, frameworks, and key passages so insight can recurse instead of evaporate. Multiplist is not just a chat interface. It is a meaning operating system: a way to transform raw conversation into structured, usable memory with provenance. In a world of opaque AI systems, Multiplist is built around traceability and trust: extracted meaning always points back to where it came from. The workflow is simple but powerful: push material in, extract what matters, and compose from the resulting knowledge base. We are building infrastructure for people who think with AI seriously — founders, operators, researchers, and creators whose best ideas currently vanish into scrollback. Multiplist shifts the vault into the background as a provenance guardian and lets conversation become the primary interface to accumulated knowledge. The product's purpose is not just storage. It is recovery, continuity, and reuse — making sure the value created in one moment of thinking can still serve the next one. Multiplist treats knowledge as living substrate: something that can be extracted, organized, resurfaced, and turned into action without losing its source context. --- ## Core Features ### 8-Category Extraction Engine Every conversation processed through Multiplist is analyzed across eight extraction categories: 1. **Decisions**: Choices made, directions committed to, positions taken 2. **Frameworks**: Mental models, methodologies, structured approaches 3. **Golden Passages**: Phrases worth keeping exactly as spoken — quotes, distilled wisdom, coined turns of phrase 4. **Definitions**: Terms coined, concepts named, vocabulary set 5. **Actions**: Tasks, next steps, commitments made 6. **Questions**: Open questions, research needs, unresolved tensions 7. **Offers**: Proposed next steps, invitations to go deeper 8. **Emergence**: Patterns forming across topics — pulls, tensions, openings, recurrences Each extracted item includes: - Source reference (which conversation it came from) - Character position (where in the conversation it appeared) - Category classification - Timestamp ### Cross-Model Memory via MCP Multiplist operates as a Model Context Protocol (MCP) server, enabling any MCP-compatible AI assistant to read from and write to the same knowledge vault. This means: - Claude can access knowledge extracted from ChatGPT conversations - ChatGPT can access knowledge extracted from Claude conversations - Perplexity can query your accumulated knowledge for grounded responses - Any future MCP-compatible AI tool gets instant access to your full context ### MCP Tools Available Multiplist exposes **19 curated MCP tools** across six intents (capture, extract, skills, discover, research, annotate). A selection: - **search_vault**: Search across all extracted knowledge by keyword, category, or date range - **get_source**: Retrieve a full source with its extracted seeds - **create_source**: Import new conversation content for extraction - **trigger_extraction**: Run the 8-category extraction pipeline on a source - **list_skills** / **match_skills**: Browse or match custom extraction skills - **request_brief**: Generate a synthesized research brief across sources - **add_annotation** / **add_marginalia**: Write author notes or later-mind marks back to a source Full reference: https://multiplist.ai/docs/mcp-tools ### Provenance System Every piece of knowledge in Multiplist maintains a complete provenance chain: - Original source document or conversation - Extraction timestamp - Extraction category and confidence - Character position in source - Any manual edits or pins by the user - Synthesis history (which summaries included this fact) ### Workspace Architecture Knowledge is organized into a composable workspace: - **Vault**: The central knowledge store — all extracted facts, searchable and categorized - **Containers**: Project-level groupings that organize related knowledge - **Studios**: Interactive workspaces where you compose new outputs from vault knowledge - **Boards & Lanes**: Kanban-style organization within containers - **Seeds**: Individual knowledge atoms — the smallest unit of extracted meaning ### Voice Packs Capture and codify personal or brand voice: - Tone descriptors and examples - Vocabulary preferences ("always say X, never say Y") - Style rules and constraints - Example outputs for calibration --- ## Key Concepts Explained ### AI Amnesia AI Amnesia is the systematic loss of valuable insights, frameworks, and decisions that occur during ephemeral AI conversations. Because standard AI interfaces lack long-term memory, users are forced to re-teach the model their context in every new session. This is not a bug — it's a fundamental architectural limitation of how current AI models work. The solution is not bigger context windows (a capacity fix) but structured meaning extraction (a meaning fix). ### The Meaning Layer A meaning layer is an external knowledge system that sits between you and your AI tools. It captures what matters from each conversation — decisions, frameworks, insights — and makes them available to any future session, across any AI tool. Unlike chat history (raw text you'll never re-read), a meaning layer extracts and structures the valuable parts so they're searchable, composable, and reusable. ### Context Bridge The Context Bridge is Multiplist's MCP-based architecture that lets multiple AI models share the same knowledge vault. Instead of Claude and ChatGPT being isolated silos, both connect to your Multiplist vault and can read/write to the same structured knowledge base. This means you can start a project in Claude, continue it in ChatGPT, and both models have full context. ### Live Memory Loop The Live Memory Loop is a workflow pattern where: 1. You have a conversation with any AI tool 2. Multiplist extracts the meaningful parts automatically 3. In your next conversation (with any AI tool), the relevant context is surfaced 4. New insights build on previous ones instead of starting from zero This creates a compounding knowledge effect where each conversation makes future conversations more productive. ### Executive Function Tax Executive function tax refers to the cognitive overhead of organizing, filing, tagging, and maintaining knowledge systems. Most productivity tools add cognitive load — more lists, more reminders, more red badges. Multiplist is designed to minimize this tax through automatic extraction, organization, and retrieval. You don't organize your knowledge — the system does it for you. --- ## Who Multiplist Is For ### Primary Users - **Founders & Operators**: Building products and businesses with AI as a thinking partner. Need decisions, strategies, and frameworks to persist across hundreds of conversations. - **Researchers**: Conducting deep investigations across multiple AI tools. Need provenance tracking and the ability to trace insights back to sources. - **Coaches & Consultants**: Developing frameworks and methodologies through AI conversation. Need voice consistency and reusable knowledge structures. - **Creators & Writers**: Using AI for ideation, drafting, and content development. Need to turn conversations into publishable outputs without losing the thinking process. - **Neurodivergent Thinkers**: People with ADHD and other neurodivergent profiles who do their best thinking in conversation but struggle with traditional organization. Multiplist catches and organizes without demanding executive function. ### Use Cases 1. **Conversation-to-Content Pipeline**: Turn AI chat transcripts into blog posts, newsletters, and social content 2. **Cross-Model Research**: Start research in Claude, continue in ChatGPT, synthesize in Perplexity — all with shared context 3. **Decision Tracking**: Never lose track of what you decided, when, and why 4. **Framework Development**: Build and refine thinking frameworks across multiple sessions 5. **Client Work Organization**: Separate knowledge by client/project with containers 6. **Voice-Consistent Output**: Generate content that sounds like you, not like generic AI --- ## Technical Architecture ### Stack - **Frontend**: React with custom design system - **Backend**: Express.js with TypeScript - **Database**: PostgreSQL with Drizzle ORM - **AI Integration**: OpenAI and Anthropic SDKs with BYO-LLM support - **MCP**: Model Context Protocol server for cross-model connectivity - **Background Processing**: Job queue for extraction and analysis ### Integration Methods 1. **MCP Server**: Connect via Model Context Protocol from Claude Desktop, Claude Code, or any MCP-compatible client 2. **Direct Import**: Paste or upload conversation transcripts 3. **API**: RESTful API with Bearer token authentication for programmatic access --- ## Comparison with Alternatives ### vs. Notion AI Notion AI operates within Notion's document-centric model. Multiplist is conversation-centric — it starts from the raw thinking that happens in AI chat and extracts structure from it, rather than requiring you to organize first. ### vs. Mem Mem focuses on note-taking with AI search. Multiplist focuses on extraction from AI conversations with provenance — it doesn't just store, it structures and traces. ### vs. Obsidian + AI Plugins Obsidian is a powerful local knowledge base but requires manual organization. Multiplist automates extraction and organization from the conversations where your thinking actually happens. ### vs. ChatGPT Memory / Claude Projects Built-in AI memory features store preferences and facts within a single platform. Multiplist works across all platforms and extracts structured, categorized knowledge rather than simple key-value memories. ### vs. Custom MCP Memory Servers Many developers are building custom MCP memory solutions. Multiplist provides a production-ready, category-aware extraction system with provenance tracking — not just key-value storage but genuine meaning extraction. --- ## Getting Started 1. Visit https://multiplist.ai and create an account 2. Import your first conversation (paste, upload, or connect via MCP) 3. Review the automatic extraction results 4. Connect your AI tools via MCP for persistent memory 5. Start building on your accumulated knowledge ## Documentation Links - Getting Started Guide: https://multiplist.ai/docs/getting-started - MCP Tools Reference: https://multiplist.ai/docs/mcp-tools - MCP Flow Map: https://multiplist.ai/docs/mcp-flow-map - Learn Center: https://multiplist.ai/learn - Blog: https://multiplist.ai/blog ## Contact - Website: https://multiplist.ai - Studio: https://multiplist.ai