Vector Memory
Give your coding agent semantic save-and-recall over project documents using a Redis vector store instead of stuffing everything into the context window.
Overview
vector-memory is an MCP server for the Build phase that stores and semantically recalls project documents through a Redis vector store using natural language.
What is this MCP server?
- Semantic document memory backed by Redis vector search
- Save and recall files with natural language queries from the agent
- PyPI package mcp-server-vector-memory 0.4.4 on stdio transport
- Registry manifest version 0.4.4 aligned with package version
- Pairs with long-running Claude Code repos that accumulate specs and notes
- Package and registry version 0.4.4
- PyPI identifier: mcp-server-vector-memory
- Transport: stdio
What problem does it solve?
Agents forget prior files and specs unless you manually re-paste them, which burns tokens and breaks continuity across coding sessions.
Who is it for?
Solo builders running multi-day agent sessions on one repo who already use or can run Redis locally or in dev.
Skip if: Teams needing enterprise document compliance, multi-tenant isolation, or memory without running and securing Redis yourself.
What do I get? / Deliverables
After setup, your agent can save documents to Redis-backed memory and pull the right chunks back with conversational queries while you keep building.
- Working save and semantic recall MCP tools over your indexed documents
- Reduced re-pasting of long specs into agent chats
- A dev-only memory layer you can tune before any production RAG service
Recommended MCP Servers
Journey fit
Memory layers are added while you build agent features and local dev infrastructure, before you optimize production observability. Natural-language file recall is core agent-tooling: it extends what the model can see across sessions without being a standalone frontend or REST product.
How it compares
Redis vector memory MCP, not a hosted RAG SaaS or generic notes skill.
Common Questions / FAQ
Who is vector-memory for?
Developers using Claude Code or Cursor who want persistent semantic recall of repo docs and notes via MCP tools.
When should I use vector-memory?
Use it while building agent-heavy workflows where the same specifications and files must be retrieved across many sessions without full context reloads.
How do I add vector-memory to my agent?
Run Redis with vector search support, install mcp-server-vector-memory from PyPI, register the stdio MCP server in your client config, and verify save/recall tools appear in the tool list.