cems-recall

name	cems-recall
description	Search memories for relevant past context, decisions, and patterns

Search Memories

Use the memory_search MCP tool to find relevant memories.

Usage

When you need context before starting work, or the user asks to recall something:

1. Detect the current project from the working directory:
   • Run git remote get-url origin to extract org/repo format
   • SSH: [e-mail]:org/repo.git → org/repo
   • HTTPS: https://github.com/org/repo.git → org/repo
   • Pass this as the project parameter to boost same-project results
2. Formulate a natural language query
3. Call memory_search with appropriate parameters (always include project)
4. If any results are truncated, use memory_get to fetch the full content
5. Use the results to inform your work

MCP Tool Call

{
  "tool": "memory_search",
  "arguments": {
    "query": "authentication patterns in this project",
    "scope": "both",
    "max_results": 10,
    "max_tokens": 4000,
    "enable_graph": true,
    "enable_query_synthesis": true,
    "project": "org/repo"
  }
}

Fetching Full Content

When search results have has_detailed: true (distilled memory) or are truncated (content ends with ...), fetch the full document:

{
  "tool": "memory_get",
  "arguments": {
    "memory_id": "the-memory-id-from-search-result"
  }
}

The response includes content_detailed (original full text) when the memory has been distilled to a summary.

Parameters

Parameter	Default	Purpose
query	(required)	Natural language search query
scope	"both"	"personal", "shared", or "both"
max_results	10	Maximum results (1-20)
max_tokens	4000	Token budget for results
enable_graph	true	Include related memories via graph traversal
enable_query_synthesis	true	Expand query with LLM for better retrieval
raw	false	Debug mode: bypass relevance filtering
project	(auto-detect)	Project ID (org/repo) — always pass this to boost same-project results

When to Search

• Before coding: Search for relevant patterns, conventions, preferences
• Before debugging: Search for similar issues and solutions
• Before decisions: Search for past architectural choices
• Before code review: Search for team conventions

Search Tips

• Always pass project — auto-detect from git remote to filter cross-project noise
• Use natural language: "how do we handle authentication" > "auth"
• Be specific: "Python backend database conventions" > "conventions"
• The system uses semantic matching, not just keywords
• Results include relevance scores and time decay ranking
• Use raw: true to debug retrieval when results seem wrong