Memory Hive
Local-first shared memory for AI agents where nothing is stored until you approve it
- Pricing
- Freemium
- Platforms
- macOS, iOS, CLI
- Runs
- Locally
- Version
- 0.1.9
- Listing updated
- 1 day ago
About Memory Hive
Memory Hive is a freemium MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.
Memory Hive, now called Apiary Memory, keeps one memory store on your own computer that several AI assistants read from. The key rule is that agents can only propose what to remember. A proposal lands in a candidate queue, and nothing enters the confirmed memory until you approve it, in the app or with the pip package's apiary-memory-cli list and approve commands.
The pip package is the open-source MIT integration layer, in plain Python, using a local SQLite file. It searches by words only, whereas the closed-source macOS app from version 1.5.0 adds a local vector model fused with keyword ranking so you can search by meaning. The README is explicit that the pip package has no GUI and that there is no Windows desktop app. The product's site lists six tools: recall, propose, manage, get_context, say and board, with a board for several AIs to hand off work.
The site describes no account, no cloud and no telemetry, with passwords and keys going into an encrypted local area. The macOS and iOS apps are free on the App Store with a free basic tier and subscriptions for the full feature set.
Install
Run in a terminal
claude mcp add hive-memory --env HIVE_AGENT_TOKEN='<HIVE_AGENT_TOKEN>' -- uvx memory-hive-mcpAdd to claude_desktop_config.json
{
"mcpServers": {
"hive-memory": {
"command": "uvx",
"args": [
"memory-hive-mcp"
],
"env": {
"HIVE_AGENT_TOKEN": "<HIVE_AGENT_TOKEN>"
}
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"hive-memory": {
"command": "uvx",
"args": [
"memory-hive-mcp"
],
"env": {
"HIVE_AGENT_TOKEN": "<HIVE_AGENT_TOKEN>"
}
}
}
}Add to .vscode/mcp.json
{
"servers": {
"hive-memory": {
"type": "stdio",
"command": "uvx",
"args": [
"memory-hive-mcp"
],
"env": {
"HIVE_AGENT_TOKEN": "<HIVE_AGENT_TOKEN>"
}
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"hive-memory": {
"command": "uvx",
"args": [
"memory-hive-mcp"
],
"env": {
"HIVE_AGENT_TOKEN": "<HIVE_AGENT_TOKEN>"
}
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"hive-memory": {
"command": "uvx",
"args": [
"memory-hive-mcp"
],
"env": {
"HIVE_AGENT_TOKEN": "<HIVE_AGENT_TOKEN>"
}
}
}
}Configuration
HIVE_MODEauto (default) asks the Hive app and falls back to the local store only when the app is unreachable; local never touches the network; remote refuses to fall back.HIVE_DBPath to the local memory database. Defaults to ~/.memory-hive/hive.db. Only used when the local store answers.HIVE_AGENT_IDAgent id issued by your own Hive app. Only needed for calls the app serves (HIVE_MODE=auto with the app running, or remote). Not required to use the local store.HIVE_AGENT_TOKENAccess token issued by your own Hive app when the agent is registered. Only needed for calls the app serves.HIVE_MCP_URLHive app MCP endpoint. Defaults to http://127.0.0.1:8768/mcp
Tools this server provides (6)
Taken from the project documentation; the running server may differ.
- recall
- Searches confirmed memory.
- propose
- Proposes a candidate memory for approval.
- manage
- Manages stored memory.
- get_context
- Returns context for the agent.
- say
- Posts a message to the collaboration board.
- board
- Shows the collaboration board.
What it does
- Propose, then approve
- Agents submit candidates and only you can confirm them into memory.
- Local SQLite store
- Memory lives in one file on your disk, by default under ~/.memory-hive.
- Shared across AIs
- Any MCP client can read the same memory, so switching assistants keeps context.
- Human CLI
- apiary-memory-cli lists, approves and searches pending candidates.
- Semantic search in the app
- The macOS app adds a local vector model with keyword fusion, while the pip package is keyword-only.
- Collaboration board
- Several AIs can hand off work with tasks waiting for your decision pinned on top.
Good for
- Keep project background once and reuse it across Claude, Cursor and Codex
- Let an agent suggest facts to remember while you review them
- Search past notes offline on your own machine
- Move to a new AI without re-explaining yourself
- Keep sensitive details such as keys in a local encrypted area
Pricing
Freemium
Free app basics with paid subscription tiers; prices not shown. MCP layer is MIT-licensed
Prices change. Check hello-fengchao.com before you buy.
Memory Hive FAQ
Can an agent write directly to my memory?
No. Agents can only propose candidates, and nothing is stored until you approve it.
Does it send data to the cloud?
The README says it has no account, no cloud and no telemetry, and the server runs on your own machine.
Is it fully open source?
Only the MCP integration layer is MIT-licensed. The macOS and iOS apps are closed source.
Is there a Windows version?
The README says there is no Windows desktop app, but the pip package runs on any OS with Python 3.8 or newer.
Does the pip package search by meaning?
No. It searches lexically only. Semantic search is in the macOS app from 1.5.0.
Is Memory Hive free?
Memory Hive has a free plan, with paid plans for more usage or features.
How do I add Memory Hive to Claude Code?
Run this command in your terminal: claude mcp add hive-memory --env HIVE_AGENT_TOKEN='<HIVE_AGENT_TOKEN>' -- uvx memory-hive-mcp
What are the best alternatives to Memory Hive?
The closest alternatives to Memory Hive are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and note1.
Alternatives to Memory Hive

Codebase Memory
MCP serverLocal MCP server that builds a code knowledge graph for AI coding agents
Runs locally- Open source
- 48.7K GitHub stars
- Knowledge and memory

XERJ
MCP serverLocal search engine for AI agents with cited answers over your code and docs
Runs locally- Open source
- 19.2K GitHub stars
- Knowledge and memory

Claude Code Ultimate Guide
MCP serverMCP server for searching the Claude Code Ultimate Guide from any MCP client
Runs locally- Open source
- 6.5K GitHub stars
- AI models and LLM tooling

Nautilus Compass Memory
MCP serverLocal-first long-term memory layer for AI agents, served over MCP
Not checked yet- 5 tools
- Remote
- Open source
- 1.3K GitHub stars
- Knowledge and memory

note1
MCP serverMCP server giving AI assistants search and summaries across your note1 meetings
- 4 tools
- Remote
- Free trial
- 1 GitHub star
- Productivity
MedXpert 图书馆连接器(通用版)
MCP serverLocal MCP server that searches and answers questions over your own Markdown wiki
Runs locally- 4 tools
- Open source
- Knowledge and memory