Memories

Memories are the foundation of Kyew. When you tell your AI to remember something, it's saved permanently and available in every future conversation — across all your AI apps and devices.

What is a Memory?

A memory is something you've told your AI to remember, stored with context so it can be found later. Each memory has:

• What to remember — the core information ("Sarah's email is sarah@company.com")
• Topic — what area it relates to ("client-reporting", "team-processes", "personal")
• Category — what kind of information ("preference", "process", "contact")
• Outcome — whether something worked or not

Your AI fills in these details automatically based on your conversation — you just say "remember that..." and it handles the rest.

Storing Memories

Just tell your AI to remember something:

"remember that weekly reports go to sarah@company.com every Friday"

Your AI saves it with context:

"remember that our team uses the #updates Slack channel for announcements"

Adding More Detail

You can specify a topic area for better organization:

"remember in the client-reporting topic that we use the blue brand template for all client decks"

What Kyew Stores

Each memory includes:

• What to remember — the core information
• Topic — what area it relates to
• Category — what kind of information it is
• Outcome — whether something worked or not

Recalling Memories

Ask your AI to recall what it knows:

"what do you remember about my reporting workflow?"

It searches your saved memories and returns what's relevant.

Search Everything at Once

"recall everything about client reporting"

This searches across both memories and saved skills in a single request.

Filter by Topic

"recall memories about client-reporting"

Limit Results

"recall the 5 most relevant memories about meetings"

Correcting Memories

If a memory has an error, just tell your AI:

"correct that memory — Sarah's email is actually sarah@newcompany.com"

The old version is kept for history, and a corrected version replaces it.

Forgetting Memories

If a memory is outdated:

"forget the memory about our old standup schedule"

The memory won't appear in searches anymore, but it's kept in your history for reference.

Memory Best Practices

Be Specific

Specific memories are more useful than vague ones:

# Good
"remember that reports go to Sarah on Fridays"

# Better
"remember that weekly client reports go to sarah@company.com every Friday
by 5pm, using the blue brand template with data from Linear"

Include Context

Context helps your AI find the right memories later:

# Minimal context
"remember that meetings are on Tuesdays"

# Rich context
"remember in the team-processes topic that our team standup is every
Tuesday at 9am on Zoom, and notes go in the #standup Slack channel"

Use Consistent Topics

Group related memories under the same topic name:

# Consistent — Kyew finds patterns easily
topic: "client-reporting"
topic: "client-reporting"
topic: "client-reporting"

# Inconsistent — harder to connect the dots
topic: "reports"
topic: "client-work"
topic: "reporting"

Record What Didn't Work Too

Knowing what failed is just as valuable:

"remember that sending the report as a PDF attachment didn't work —
Sarah prefers a direct link to the Google Doc"

How Memories Become Skills

Memories are the building blocks for skills — reusable workflows your AI can apply automatically:

1. Save memories around a topic
2. Kyew notices patterns (3+ related memories)
3. A skill is generated from those patterns
4. You approve it and your AI applies it going forward

The more specific and consistent your memories, the better the skills Kyew creates.

API Reference

For the full technical reference including all parameters and options, see Memory Tools API.