HiveMind (projectmentor) CLI
Shared memory for AI agents across machines: a fact one agent writes shows up for the others with no server in between
- 73 ★ on the repository
- Last commit
- AGPL-3.0
Extended access
Why this level
- The curl and bash installer sets up Tailscale, clones the repository and starts a sync service on the machine
- Agent notes replicate between your devices over Tailscale and may contain work data
Install
- Works in
Manual install
curl -fsSL https://raw.githubusercontent.com/projectmentor/hive-mind/main/scripts/installer/install.sh | bashThen hive-mind install. The installer sets up Tailscale and the sync service.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The curl and bash installer sets up Tailscale, clones the repository and starts a sync service on the machine
- Agent notes replicate between your devices over Tailscale and may contain work data
This is third-party code. Review the repository files before installing.
What it does
HiveMind keeps agent notes as a shared hive and syncs it across devices. The `hv remember` command writes a fact with tags, and `hv search` looks through the notes on the current machine without calling out to the internet. Results sort by trust, importance, utility or recency. Every fact records its author and time, and old lines stay. Trust in a fact grows when different agents agree on their own, and repeating it from one agent does not raise the score. Agents on one device count for less than agents on different devices. If two notes disagree, you see both. A decision can name the facts it used and, once its outcome is recorded, gets a score while those facts get a utility score. Sync runs over Tailscale and passes only new notes between devices. A new device can read at first and can write after you admit it with `hive-mind group admit`. Without a network an agent keeps working and the notes merge when the devices next meet. There is no database or cloud account to set up, and a backup is a copy of the files. It connects to Claude Code through the `hive-memory` skill with hooks, to Hermes through a memory plugin and to Claude Desktop through a local MCP server.
Who it is for. People who run agents on several of their own machines, including an Android phone, and want one agent findings not to get lost for another.
Good fit when
- An agent on a laptop found a payments API rate limit and an agent on a server should know it without asking again
- A planner, a builder and a checker run as three agents and keep shared project notes
- Two notes disagree and you need both to survive the merge
- You want memory without a cloud service, a database and an account
Not a fit when
- You need memory for one application with a service in the middle: Mem0 covers that
- You need a stable agent contract: the project is alpha and contract 4.2 changes between minor versions
- An agent runs on one machine only with no recurring notes, and shared memory is not needed
Example request
Remember in the hive that the payments API is limited to 100 requests per second, and add the tags api and payments.Limitations
The project is alpha: the agent contract is now 4.2 (the `hv version` command) and changes between minor versions. The installer runs as `curl ... | bash`, installs Tailscale with authentication, clones the repository to `~/projects/hive-mind` and starts a sync service under systemd, launchd or runit. Platforms: Android through Termux, Linux, macOS and Windows 11 with WSL2 and systemd enabled.
Tailscale is needed only to sync several machines. On Android it is a separate app that the installer cannot manage from Termux, and a phone behind the userspace VPN mode syncs through outbound connections every five minutes. The license is AGPL-3.0, the last commit is from October 18, 2026.
How to disable. Stop and disable the sync service with your OS tools (systemd, launchd or runit), then delete the `~/projects/hive-mind` directory and the local HiveMind database. Remove the `hive-memory` skill and hooks from the Claude Code configuration.
FAQ
How do I install HiveMind?
Run `curl -fsSL https://raw.githubusercontent.com/projectmentor/hive-mind/main/scripts/installer/install.sh | bash`, then `hive-mind install`. The installer asks whether to join an existing hive or start your own, and on the first machine you just press Enter.
How do I add a second machine to the hive?
Run `hive-mind invite` on the first machine and it prints a line. Paste it when running `hive-mind install` on the second machine. The new device starts writing after you admit it with `hive-mind group admit`.
How is HiveMind different from Mem0?
Mem0 keeps memory for an application, usually with a service in the middle. HiveMind keeps the agents own notes on your machines and shows disagreements between them.
Which agents does HiveMind work with?
Claude Code, Hermes and Claude Desktop through a local MCP server. Any agent that can run a command can use `hv`.
Do I need internet and a cloud account?
The install and git clone need internet, after that an agent works offline. No cloud account or database is needed, and Tailscale is required only to sync between machines.
Discussion
No questions or reviews yet.
Related
CLI toolsShared cloud memory for a team's agents: stores session traces and turns repeated patterns into skills
External memory for Hermes, Claude Code, Codex and Cursor: session archive, recall layers and relevant context injected back
Persistent agent memory in a folder of linked markdown notes, so Claude Code remembers a project across sessions
Shared memory for Claude Code, Codex and Cursor: a decision made in one session is recalled in another tool or on another machine.
Long-term memory for AI agents: a skill and the mindmemos command store facts from conversations and recall them in later sessions
Local persistent memory for coding agents: hooks record every session, and Claude Code, Codex or Cursor recall past decisions through 54 MCP tools
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| HiveMind (projectmentor) (this card) | no | 73 | |
| Mem0 | no | 68,887 | stores memory for an application, usually through a service, while this one keeps the agents own notes on your machines with no server |
| Claude-Mem | no | 116,457 | keeps Claude Code session context on one machine, while this one syncs the memory of several agents across devices |
README in short
The description explains how agents share memory without a central server: trust grows through independent agreement, disagreements stay visible and search is local. It then covers recommendations for the planner, builder and checker roles, installation and platform requirements.