Skip to main content

​
Purpose

mem_get retrieves complete information about specific observations. Use this after mem_search to get the full details you need. This tool is designed for efficiency:
  • Only retrieve observations you actually need
  • Costs tokens only for what you request
  • Supports batch retrieval of multiple observations

​
Parameters

​
ids
array
required
Array of observation IDs to retrieve (from mem_search, max: 20)Example: [142, 156, 203]

​
Response Format

Returns full details for each observation: Example response: 
**[142] bash** — 2026-03-01 14:32
Summary: Started development server on port 3000
Files: server.ts, package.json
Concepts: development, server, port-configuration
Input: bun run dev
Output: Server started successfully
> Local: http://localhost:3000
> Network: http://192.168.1.5:3000

---

**[156] edit** — 2026-03-02 09:15
Summary: Added authentication middleware to API routes
Files: api/routes.ts, middleware/auth.ts
Concepts: authentication, middleware, api-security
Input: {"filePath": "/api/routes.ts", "oldString": "app.get('/api', handler)", "newString": "app.get('/api', authMiddleware, handler)"}
Output: Successfully updated routes.ts with authentication middleware. Applied to 5 routes.

---

**[203] read** — 2026-03-03 16:20
Summary: Reviewed database connection configuration
Files: config/database.ts
Concepts: database, configuration, connection-pool
Input: config/database.ts
Output: [file contents showing connection pool settings, max connections: 10, timeout: 5000ms]
Each observation includes:
  • ID and tool name (header)
  • Timestamp (YYYY-MM-DD HH:MM)
  • Summary (compressed description)
  • Files (referenced files, if any)
  • Concepts (extracted concepts/tags)
  • Input (tool input, truncated to 200 chars)
  • Output (tool output, truncated to 500 chars)

​
Usage Example

Get a single observation: 
{
  "ids": [142]
}
Get multiple observations at once: 
{
  "ids": [142, 156, 203]
}
Typical workflow: 
// Step 1: Search for relevant context
{
  "tool": "mem_search",
  "query": "authentication",
  "limit": 10
}

// Step 2: Get full details for interesting results
{
  "tool": "mem_get",
  "ids": [142, 156, 203]
}

​
Tips

  • Use after search: Always use mem_search first to find relevant IDs
  • Batch requests: Request multiple IDs at once instead of making separate calls
  • Token efficiency: Only request observations you actually need — each observation adds to context
  • Max limit: You can request up to 20 observations per call
  • Empty results: If none of the IDs exist, you’ll get “No observations found for those IDs.”
  • Truncated content: Input/output are truncated for token efficiency. Use mem_export with include_raw for full data

Build docs developers (and LLMs) love

Get started for freeTalk to us