Memory as a Service · Now in beta

Your AI agents
remember everything.

Mnemoe is a persistent, graph-structured knowledge base that your AI tools can read from and write to. Every insight from every session — remembered and reused.

Works with any AI tool

ClaudeCursorChatGPTGeminiGitHub CopilotAny HTTP client
~/.claude/CLAUDE.md

# Mnemoe

MNEMOE_API=https://api.mnemoe.com/api/v1

MNEMOE_KEY=mem_••••••••••••••••••••

# At the start of every session:

GET $MNEMOE_API/instructions

Authorization: Bearer $MNEMOE_KEY

# Follow the returned instructions. That's it.

Agent connected · 14 articles loaded
Markdown

How it works

Three steps. No infrastructure.

Works with the tools you already use.

01

Add a snippet to your AI tool

Paste a small config block into CLAUDE.md, a Cursor rule, or your system prompt. Your agent now knows where to look.

02

Agents retrieve context automatically

Before answering, your agent searches the KB for relevant articles. Past decisions and lessons — instantly available.

03

You approve what gets remembered

After a session, the agent proposes what to save. You review it in the Suggestions queue — nothing is written without you.

Features

Everything your agents need

Semantic search

pgvector-powered similarity search returns the most relevant knowledge for any query — not just keywords.

Knowledge graph

Articles link to each other. Visualise your entire KB as an interactive graph to spot gaps and clusters.

Human-in-the-loop writes

Agents can only suggest. Every write goes through a review queue — you stay in control.

Usage insights

See which articles get surfaced most, which queries trigger them, and what your agents actually use.

API key per agent

Issue separate keys for Claude, Cursor, scripts, and CI. Revoke any without affecting the others.

Works everywhere

A simple REST API. If your tool can make an HTTP call, it can use Mnemoe. No SDKs required.

Integrate

Live in 30 seconds

Add this to ~/.claude/CLAUDE.md and you're done.

claude / cursor / any agent

# 1. fetch your instructions once

GET $MNEMOE_API/instructions

# 2. search context before answering

POST $MNEMOE_API/context

{ "query": "...", "limit": 5 }

# 3. suggest learnings at session end

POST $MNEMOE_API/suggestions

{ "conversation": "...", ... }

Claude Code
Add to ~/.claude/CLAUDE.md
Cursor
Add to .cursor/rules
System prompt
Works with any model via API
Custom scripts
REST API — any HTTP client

“Every AI session used to start from zero. Mnemoe means my agents walk in already knowing our architecture, past decisions, and open questions. It's the persistent brain I never had.”

A

Ari Nakamura

Building with Claude & Cursor every day

Get started today

Give your AI
a memory.

Free to get started. No credit card required.

Create your knowledge base →

Already have an account? Sign in