Persistent memory for your AI · MCP-native

A knowledge base your AI builds for you.

Connect Lore once. Talk to your AI like normal. When something matters — a decision, an insight, a plan — it files it into a structured database that's yours to read and edit.

Get early access See how it works

ClaudeOpenClawany MCP client

Connect. Talk. Own.

Three verbs. No new workflow. No tagging by hand. No filing. You already have the conversations — Lore gives them somewhere to live.

01

Connect

Paste one URL.

Drop the Lore URL into Claude's connector settings (or OpenClaw, or any MCP client) and sign in once. Takes under a minute.

uselore.app/mcp

02

Talk

Have your usual conversations.

When something's worth keeping, the AI saves it — typed, tagged, under the right naming convention. It tells you what it did.

Created: DECISION — Postgres over SQLite

03

Own

Read anything. Edit anything.

Your knowledge base is a real database on your side of the wall. Browse, correct, archive. No black box.

data/users/you/kb.sqlite

What your AI can do

Nine verbs, one knowledge base.

Lore is a small, careful toolset. The AI reads these descriptions once and knows how to work with your KB — that's the whole product.

Orient

get_manual

Load your taxonomy, conventions, and counts at the start of every session.

Shape

configure

Set up or adjust entry types, tags, statuses, naming rules.

Write

create_entry

Save a new entry — after searching first, so nothing duplicates.

Revise

update_entry

Append a dated update, or carefully replace stale content.

Read

get_entry

Pull the full content of a specific entry by id.

Skim

list_sections

Browse section headings without their content, so the AI can narrow in before it reads.

Focus

get_section

Read one section of an entry rather than the whole thing.

Browse

list_entries

Filter by type, status, tag, or date to see what's in the drawer.

Find

search

Search across every entry — by exact words, or by meaning. Ranked by relevance so the AI can answer "do I have anything about X?" without guessing.

Examples

No two drawers look alike.

On first connection, your AI interviews you and proposes a taxonomy shaped around what you actually do. Here's what other drawers look like.

Student

Lecture Essay Study plan Research Reflection

23 entries · 4 tags

Founder

Meeting Investor update Hiring decision Strategy Build spec

147 entries · 11 tags

Creative

Mood board Client brief Inspiration Project log Reflection

64 entries · 7 tags

You

Sign up and your AI walks you through setup. Five minutes of conversation, and you'll have a fully-shaped drawer waiting.

Begin

Trust

Your data, on your side of the wall.

i.

Isolated

Every user gets their own encrypted SQLite database. Nothing shared between users, nothing used for training. With search by meaning on — the standard — section text goes to VoyageAI to be indexed, and each search by meaning sends the query and the candidate sections (with their entry's name, type and description) to be ranked. Nothing else, to nobody else.

ii.

Yours

Everything your AI writes belongs to you. Export all of it any time, as one JSON file, and take your Lore to another tool.

iii.

Transparent

Read every word your AI has written about you. Edit when it drifts. No hidden state, no "memory" you can't inspect.