recap vs digest vs search

MemNexus has four main ways to retrieve information, and choosing the right one makes a real difference in what you get back and how fast. Here's when to use each.

The four methods

Method	Speed	Output	Best for
memories search	Fast	Raw matching memories	Finding a specific memory or fact
memories recap	Fast	Raw memories grouped by conversation	"What did I work on recently?"
memories digest	Slow	AI-synthesized narrative	"Give me an overview of X"
conversations summary	Medium	AI summary of one session	Deep-diving a specific work session

memories search — targeted lookup

Use search when you're looking for something specific: a decision, a configuration value, a previous solution to a known problem.

# Find the specific memory about rate limiting
mx memories search --query "rate limit configuration decision"

# Find what went wrong last time you touched this service
mx memories search --query "auth service outage root cause"

# Quick lookup with --brief (top 5, compact format)
mx memories search --query "redis connection pool" --brief

Search is the right choice when you can describe what you're looking for in a few words. If you know the answer exists somewhere in your memories and you need to retrieve it, search finds it.

Add --timeline for investigation: When you're trying to understand how something evolved over time — how a decision changed, how a problem developed — timeline mode sorts by date and shows state indicators ([CURRENT], [SUPERSEDED], [CONTRADICTED]):

mx memories search --query "auth architecture" --timeline

Add --mode facts for specific knowledge: The facts search mode looks specifically at extracted structured facts, which is useful for precise lookups:

mx memories search --query "rate limit" --mode facts

memories recap — recent work, raw

Recap returns raw memories grouped by conversation, in reverse chronological order. It's the fastest way to see what you've been working on.

# What did I work on in the last day?
mx memories recap --recent 24h

# What happened this week on the billing feature?
mx memories recap --recent 7d --query "billing"

Use recap for:

• Standup prep — See what you worked on yesterday without having to reconstruct it
• Resuming after a break — Get back up to speed on where you were
• Quick context for a new session — See the last few conversations before diving in

Recap is raw — it shows you the memory content directly, grouped by conversation, without synthesis. This is a feature: it's fast and you see exactly what was saved, not an AI's interpretation of it.

memories digest — synthesized overview

Digest uses an AI to read your relevant memories and write a narrative summary. It's slower but returns something more readable for topics with lots of memories.

# Get an overview of your authentication work
mx memories digest --query "authentication" --digest-format structured

# Status report for a standup
mx memories digest --query "sprint work" --recent 7d --digest-format status-report

# Timeline of how a feature developed
mx memories digest --query "rate limiting feature" --digest-format timeline

The three digest formats:

• structured — Organized sections: what was done, key decisions, current state, next steps
• status-report — Standup-style: what's done, what's in progress, blockers
• timeline — Chronological narrative: what happened and when

Use digest when:

• You're unfamiliar with a topic and need context before diving in — a feature someone else built, or something you haven't touched in months
• You need a narrative, not a list — preparing a handoff, writing a postmortem, summarizing work for someone else
• You have many memories on a topic and search results feel overwhelming

conversations summary — deep dive on a work session

When you know which conversation you're interested in, conversations summary gives you an AI-generated summary of that specific session.

# First, find the conversation ID
mx conversations list --recent 7d

# Then summarize it
mx conversations summary conv_abc123

# Or see the full timeline of events
mx conversations timeline conv_abc123

Use this when:

• You remember doing something specific last week but can't recall the details
• You're handing off a piece of work and want to summarize the recent history
• You want to brief someone on a debugging session without walking them through every step

Decision guide

"I need to find a specific memory about X." → mx memories search --query "X"

"I need to find where a decision came from or how it evolved." → mx memories search --query "X" --timeline

"What did I work on today/this week?" → mx memories recap --recent 24h or --recent 7d

"I'm starting a session on X — give me context." → mx memories digest --query "X" --digest-format structured

"I need a status update on X for my standup/manager/team." → mx memories digest --query "X" --recent 7d --digest-format status-report

"What happened in that debugging session last Tuesday?" → mx conversations list --recent 7d, then mx conversations summary <id>

"I need to look up a specific fact or configuration value." → mx memories search --query "X" --mode facts or mx facts search --query "X"

Combining methods

For getting up to speed on a new session, a two-step approach often works well:

# Step 1: Fast raw look at recent work
mx memories recap --recent 24h

# Step 2: Synthesized overview of the specific area you're about to work on
mx memories digest --query "payments service" --digest-format structured

Recap shows you what you were doing. Digest gives you the synthesized state of the specific thing you're about to touch.

Next steps

• Session start ritual — How to structure the beginning of each AI session
• Capturing memories that actually help — What to save and how to write it
• Search reference — Full search syntax and options