# MemWire

> Open source & enterprise-ready AI memory infrastructure layer

## Docs

- [Claude Code setup](https://memwirelabs.ai/ai-tools/claude-code.md): Configure Claude Code for your documentation workflow
- [Cursor setup](https://memwirelabs.ai/ai-tools/cursor.md): Configure Cursor for your documentation workflow
- [Windsurf setup](https://memwirelabs.ai/ai-tools/windsurf.md): Configure Windsurf for your documentation workflow
- [Health check](https://memwirelabs.ai/api-reference/endpoint/health.md): Returns the server status. Use this to verify the API is up before sending requests.
- [Add knowledge base](https://memwirelabs.ai/api-reference/endpoint/knowledge-add.md): Ingest a named knowledge base of document chunks. Chunks are embedded and indexed for semantic search and recall.
- [Delete knowledge base](https://memwirelabs.ai/api-reference/endpoint/knowledge-delete.md): Delete a knowledge base and all its chunks by kb_id.
- [Search knowledge](https://memwirelabs.ai/api-reference/endpoint/knowledge-search.md): Search knowledge base chunks by semantic similarity using hybrid dense+sparse retrieval.
- [Recall memory context](https://memwirelabs.ai/api-reference/endpoint/recall.md): Traverse the memory graph via BFS to find the most relevant context for a natural-language query. Returns a formatted string ready to inject into your LLM prompt.
- [Search memories](https://memwirelabs.ai/api-reference/endpoint/search.md): Semantic similarity search over stored memories using hybrid dense + sparse retrieval. Optionally filter by category.
- [Store memory](https://memwirelabs.ai/api-reference/endpoint/store.md): Store one or more conversation messages into the user's memory graph. Each message is embedded, classified, and indexed — ready for recall and search.
- [API reference](https://memwirelabs.ai/api-reference/introduction.md): REST API for storing, recalling, and searching memories
- [Configure OSS Stack](https://memwirelabs.ai/configuration.md): Configure the MemWire stack with your preferences for vector store, LLM, embedding model, search quality, recall tuning, and more.
- [Local development](https://memwirelabs.ai/development.md): Run the MemWire docs locally with the Mintlify CLI
- [Overview](https://memwirelabs.ai/embedders.md): Choose from the supported embedding backends and models.
- [FastEmbed](https://memwirelabs.ai/embedders/fastembed.md): Run local dense and sparse embeddings with Qdrant's FastEmbed library — no external API required.
- [Code blocks](https://memwirelabs.ai/essentials/code.md): Display inline code and code blocks
- [Images and embeds](https://memwirelabs.ai/essentials/images.md): Add image, video, and other HTML elements
- [Markdown syntax](https://memwirelabs.ai/essentials/markdown.md): Text, title, and styling in standard markdown
- [Navigation](https://memwirelabs.ai/essentials/navigation.md): The navigation field in docs.json defines the pages that go in the navigation menu
- [Reusable snippets](https://memwirelabs.ai/essentials/reusable-snippets.md): Reusable, custom snippets to keep content in sync
- [Global Settings](https://memwirelabs.ai/essentials/settings.md): Mintlify gives you complete control over the look and feel of your documentation using the docs.json file
- [Adaptive Feedback Loop](https://memwirelabs.ai/features/adaptive-feedback.md): Graph edge weights update from real LLM responses, so the memory graph improves with every conversation.
- [Graph Memory](https://memwirelabs.ai/features/graph-memory.md): How MemWire builds and traverses a token-level displacement graph to recall semantically connected memories.
- [Hybrid Search](https://memwirelabs.ai/features/hybrid-search.md): Combine dense semantic vectors and sparse SPLADE keyword vectors for higher-recall retrieval.
- [Knowledge Base](https://memwirelabs.ai/features/knowledge-base.md): Ingest documents and chunks that are searched alongside episodic memories at recall time.
- [LLM-Free Classification](https://memwirelabs.ai/features/llm-free-classification.md): Auto-tag every memory by any category and no external LLM API calls.
- [Multi-Tenancy](https://memwirelabs.ai/features/multi-tenancy.md): Isolate memory across organisations, workspaces, apps, and users with a native four-level hierarchy.
- [Overview](https://memwirelabs.ai/features/overview.md): Core capabilities that make MemWire unique.
- [Reranker](https://memwirelabs.ai/features/reranker.md): Boost search precision with a local cross-encoder that rescores candidates using full query-document attention.
- [Introduction](https://memwirelabs.ai/index.md): Enterprise-grade, self-hosted AI memory infrastructure layer. Deploy persistent AI memory on-premise or in any cloud with your own LLM and database.
- [Overview](https://memwirelabs.ai/llms.md): MemWire is model-agnostic. Use recalled memory context with any LLM provider.
- [Azure OpenAI](https://memwirelabs.ai/llms/azure-openai.md): Use Azure OpenAI deployments with MemWire memory context.
- [OpenAI](https://memwirelabs.ai/llms/openai.md): Use OpenAI GPT models with MemWire memory context.
- [Quickstart](https://memwirelabs.ai/quickstart.md): Add persistent memory to your AI agent in minutes
- [Overview](https://memwirelabs.ai/vector-databases.md): Connect MemWire to a vector store.
- [Qdrant](https://memwirelabs.ai/vector-databases/qdrant.md): Configure MemWire with Qdrant — embedded mode, local Docker server, or Qdrant Cloud.

## OpenAPI Specs

- [openapi](https://memwirelabs.ai/api-reference/openapi.json)

## Optional

- [GitHub](https://github.com/memoryoss/memwire)
- [Roadmap](https://github.com/memoryoss/memwire/blob/main/ROADMAP.md)