mem-search

name	mem-search
description	Search kimi-mneme's persistent cross-session memory database. Use when user asks "did we already solve this?", "how did we do X last time?", "what did we do yesterday?", or needs work from previous sessions.

Memory Search

Search past work across all sessions. Simple workflow: search → timeline → get.

When to Use

Use when users ask about PREVIOUS sessions (not current conversation):

• "Did we already fix this?"
• "How did we solve X last time?"
• "What happened last week?"
• "What did we do yesterday?"
• "Remind me about the auth system"

3-Layer Workflow (ALWAYS Follow)

NEVER fetch full details without filtering first. 10x token savings.

Step 1: Search — Get Index with IDs

Use the mneme_search plugin tool:

mneme_search(query="authentication", limit=20, project="my-project")

Returns: Table with IDs, timestamps, types, titles (~50-100 tokens/result)

| ID | Time | Type | Snippet |
|----|------|------|---------|
| #11131 | 3:48 PM | PostToolUse | Added JWT authentication... |
| #10942 | 2:15 PM | PostToolUse | Fixed auth token expiration... |

Parameters:

• query (string) — Search term (required)
• limit (number) — Max results, default 10, max 50
• project (string) — Project name filter
• type (string, optional) — "bugfix", "feature", "refactor", "docs", "test"
• date_from (string, optional) — YYYY-MM-DD
• date_to (string, optional) — YYYY-MM-DD

Step 2: Timeline — Get Context Around Interesting Results

Use the mneme_timeline plugin tool:

mneme_timeline(observation_id=11131, radius=5)

Returns: radius items before + center + radius items after in chronological order.

Parameters:

• observation_id (number, required) — Observation ID to center around
• radius (number, optional) — Items before/after, default 5, max 20

Step 3: Get — Fetch Full Details ONLY for Filtered IDs

Review titles from Step 1 and context from Step 2. Pick relevant IDs. Discard the rest.

Use the mneme_get plugin tool:

mneme_get(ids=[11131, 10942])

ALWAYS use mneme_get for 2+ observations — single request vs N requests.

Parameters:

• ids (array of numbers, required) — Observation IDs to fetch

Returns: Complete observation objects with tool_input, tool_output, error, prompt (~500-1000 tokens each)

Examples

Find recent bug fixes:

mneme_search(query="bug", limit=20)

Find what happened last week:

mneme_search(query="", date_from="2026-04-28", limit=20)

Understand context around a discovery:

mneme_timeline(observation_id=11131, radius=5)

Batch fetch details:

mneme_get(ids=[11131, 10942, 10855])

Why This Workflow?

• Search index: ~50-100 tokens per result
• Full observation: ~500-1000 tokens each
• Batch fetch: 1 request vs N requests
• 10x token savings by filtering before fetching

Auto-Injected Context

On session start, kimi-mneme automatically injects:

• Recent project activity summary
• Recurring patterns (errors, fixes)
• Structured observations from previous sessions

You can reference this context directly without searching.