Lodekeep

Lodekeep is a hosted portable persistent memory service for AI assistants. It implements a Model Context Protocol (MCP) memory server so that ChatGPT, Claude, Grok, Gemini, Cursor, Claude Code, and other MCP clients can share one memory layer. Instead of storing isolated memories in each AI product, users can tell one assistant to remember a decision or preference and later have a different connected assistant recall it. The service provides a free tier, a live demo that does not require an account, and a simple setup path where compatible clients can be connected by pasting one URL or running one command in CLI tools. Lodekeep emphasizes user control and curation. Users can include or exclude project sources from recall with a tap, ensuring only chosen repositories or work streams feed the AI. It also offers personas—adjustable standards or 'rigor ladders' such as Enterprise Grade, Client-Ready, Startup MVP, and Research—that agents are asked to follow. Personas can be blended and given an intensity level, while a security floor prevents the selected standard from dropping below a user-defined minimum. Under the hood, Lodekeep performs semantic recall using OpenAI embeddings and vector search with keyword fallback. This means users can ask in natural language and retrieve relevant decisions, lessons, and context scoped to their own memory. The same memory is exposed through an MCP bridge and an API with scoped, revocable keys, making it suitable for always-on autonomous agents running on servers that may restart or switch models. Memorable context is restored after a restart, avoiding the need to re-establish what an agent was working on.

Category: AI/ML
Pricing: Free

Problem solved

AI assistants and agents typically have fragmented or short-lived memory: context is lost when a user switches tabs, tools, models, or sessions, and each AI stores its own separate memories. Lodekeep solves this by acting as a vendor-neutral MCP memory server connected to major assistants. Users capture a piece of information once, curate which sources it can use, and any connected AI can recall it later—while the user retains control and can export or delete memories at any time.

Use cases

Give ChatGPT, Claude, Grok, or Gemini a shared memory of preferences and decisions across tools.
Remember coding decisions in Cursor, Claude Code, or Gemini CLI so they carry across sessions.
Scope memory by project so only selected repositories or workflows influence an AI's answers.
Set persona standards such as Enterprise Grade or Startup MVP to make agents follow consistent quality/security thresholds.
Restart an always-on autonomous agent without losing context about ongoing tasks and constraints.
Connect any MCP-compatible client to one memory API with a scoped, revocable key.
Use the interactive demo or a one-URL connector to try persistent memory before creating an account.

Audience

Developers
Startups
Enterprise
Hobbyists

Pros

Vendor-neutral: works across many leading AI assistants and MCP clients
User-controlled curation and scope for what an AI recalls
Simple setup for compatible clients: paste a URL or run a one-line command
Semantic recall with natural-language search and keyword fallback
Free tier forever and export/delete capabilities

Cons

Free tier stores memories for only 30 days; longer retention requires a paid plan
Requires MCP-compatible clients or API integration for the full experience
As a newer memory provider, it is less proven than built-in memory features in some assistants
Recall quality depends on source selection, embeddings, and the connected client
Not a general-purpose knowledge base; designed for scoped persistent memory snippets

Official website: https://lodekeep.com/

Canonical LambdaBase page: https://www.lambdabase.com/tools/lodekeep

← Directory

Lodekeep

Free

★ 12 • Updated Sep 8, 2026, 03:16 AM

Lodekeep is a hosted portable persistent memory service for AI assistants.

Use Cases ▼

Give ChatGPT, Claude, Grok, or Gemini a shared memory of preferences and decisions across tools.

Remember coding decisions in Cursor, Claude Code, or Gemini CLI so they carry across sessions.

Scope memory by project so only selected repositories or workflows influence an AI's answers.

Set persona standards such as Enterprise Grade or Startup MVP to make agents follow consistent quality/security thresholds.

Restart an always-on autonomous agent without losing context about ongoing tasks and constraints.

Connect any MCP-compatible client to one memory API with a scoped, revocable key.

Use the interactive demo or a one-URL connector to try persistent memory before creating an account.

What Problem It Solves ▼

Key Features ▼

✓

Vendor-neutral: works across many leading AI assistants and MCP clients

✓

User-controlled curation and scope for what an AI recalls

✓

Simple setup for compatible clients: paste a URL or run a one-line command

✓

Semantic recall with natural-language search and keyword fallback

✓

Free tier forever and export/delete capabilities

Overview

Full Description ▼

Lodekeep emphasizes user control and curation. Users can include or exclude project sources from recall with a tap, ensuring only chosen repositories or work streams feed the AI. It also offers personas—adjustable standards or 'rigor ladders' such as Enterprise Grade, Client-Ready, Startup MVP, and Research—that agents are asked to follow. Personas can be blended and given an intensity level, while a security floor prevents the selected standard from dropping below a user-defined minimum.

Under the hood, Lodekeep performs semantic recall using OpenAI embeddings and vector search with keyword fallback. This means users can ask in natural language and retrieve relevant decisions, lessons, and context scoped to their own memory. The same memory is exposed through an MCP bridge and an API with scoped, revocable keys, making it suitable for always-on autonomous agents running on servers that may restart or switch models. Memorable context is restored after a restart, avoiding the need to re-establish what an agent was working on.