ReactiveMemory MCP Server

ReactiveMemory MCP Server gives AI assistants a persistent, queryable, locally-stored memory system backed by vector search, a temporal knowledge graph, and a vault-structured content store. It is designed to be the durable external memory layer for agents, copilots, and AI workflows that need to remember, reason over, and navigate accumulated knowledge across sessions.

It is implemented in C# on net10.0 using ModelContextProtocol 1.3.0.

Quick Install

Click to install in your preferred environment:

Note:

• These install links are prepared for the intended NuGet package identity CP.ReactiveMemory.Mcp.Server.
• If the latest package has not been published yet, use the manual source-build configuration below.

Manual MCP configuration using NuGet:

{
  "mcpServers": {
    "reactivememory-mcp-server": {
      "type": "stdio",
      "command": "dnx",
      "args": [
        "CP.ReactiveMemory.Mcp.Server@1.*",
        "--yes"
      ]
    }
  }
}

Getting started

1. Install the MCP server with the quick-install link or the dnx configuration above.
2. Restart or reload your MCP-capable client so it discovers the reactivememory_* tools.
3. Ask your assistant to call reactivememory_status to confirm the core path, current taxonomy, and fallback-safe local model status.
4. For memory-aware work, have the assistant call reactivememory_memory_automanage or reactivememory_react_to_prompt at the start of each user prompt, then use reactivememory_memory_get_relevant, reactivememory_search_relays, reactivememory_search, or reactivememory_facts_query before answering from memory.
5. After meaningful work, store durable outcomes with reactivememory_memory_add, reactivememory_add_drawer, or reactivememory_diary_write rather than saving full transcripts. Use reactivememory_memory_classify / reactivememory_memory_should_store to audit decisions, reactivememory_memory_summarise to compress long-term groups, and reactivememory_memory_prune for dry-run pruning recommendations unless apply=true is explicitly requested.

Minimal first prompt:

Call reactivememory_status, then react to this prompt with reactivememory_react_to_prompt. Search for any prior memories about this repository before planning the work.

What ReactiveMemory helps with

Without persistent memory, every AI session starts from zero. ReactiveMemory solves this by giving agents a place to:

• Store any content — code snippets, decisions, meeting notes, design facts — under an organized sector/vault taxonomy
• Search acr

... View full README on GitHub