honeycomb FAQ

Short, direct answers to what people actually ask about honeycomb.

#The basics

#What is honeycomb in one sentence?

A shared, lasting memory for your AI coding assistants, so what one of them learns is remembered by all of them, across sessions, tools, devices, and your team if you want.

#Do I need to know about databases or Deeplake to use it?

No. You install with one command, finish setup with a click in the hive dashboard, and use plain commands like honeycomb remember and honeycomb recall. The technical machinery stays out of sight.

#Which AI coding assistants does it work with?

Claude Code, Cursor, and Codex today, with Hermes, pi, and OpenClaw in progress. honeycomb plugs in underneath whichever ones you have installed, and a memory written from one is recalled by the others.

#Who makes honeycomb?

It's a collaboration between Legion Code and Activeloop. Activeloop provides Deeplake, the database honeycomb's memory lives in, and Hivemind, the open-source project honeycomb is built on. Legion Code adds the three-tier memory system, skill sharing, the self-tidying loop, and the local daemon that ties it all together.

#Privacy and data

#Where does my data live?

In your own Deeplake store, which you control and can even host in your own cloud account. The daemon on your machine is the only thing that connects to it.

#Can other people or teams see my memories?

Not unless you choose to share. Different companies, teams, and projects are kept separate at the storage layer, and inside a team the default leans private. You widen sharing on purpose, never by accident.

#Are my API keys and secrets safe?

Yes. Secrets are stored separately from memory, encrypted, and never shown to an assistant. An assistant can cause a secret to be used, for example to call a service, without the value ever passing through it.

#Does honeycomb send my code or prompts anywhere?

The only outbound traffic is signing in with Deeplake and, if you allow it, anonymous product-usage counts to help the makers understand adoption. That usage signal never includes your code, prompts, memories, file paths, or names, and you can turn it off entirely with honeycomb telemetry --show and the relevant environment variable. Your actual memories go only to the store you control.

#Can I stop honeycomb from recording?

Yes. Put it in read-only mode for a session (recall still works, nothing new is written), which is handy when you're working with something sensitive.

#Cost and performance

#Does honeycomb slow my assistant down?

No. Recording is cheap and stays out of the way, and if anything hiccups, your assistant keeps working normally. The briefing it adds at the start of a session is deliberately small.

#Does honeycomb cost money in AI model usage?

Everyday memory features, capturing, recalling, and the start-of-session briefing, don't need their own AI model or API key. Two optional extras can use a model: turning sessions into summaries and skills, and the periodic self-tidying loop. Both are opt-in, so you decide when to spend.

#Do I need an internet connection?

You need to be signed in to reach your store. The optional semantic search feature runs on a small language model downloaded once and run locally on your machine, not a cloud service.

#How is this different from a plain vector database or "RAG"?

A plain vector database stores text and hands back similar text. honeycomb does more: it keeps memory at three levels of detail so an assistant can skim then zoom in, it tidies itself over time so it gets sharper instead of noisier, it turns lessons into shareable skills, and it works across many tools and your whole team. The storage underneath, Deeplake, handles both exact lookups and meaning-based search in one place, with full version history.

#What is semantic search, and do I need it?

It's the ability to find a memory by what it means, even when you used different words than the original note. It's optional. With it on, honeycomb catches more of the "I didn't know the exact term" cases. With it off, recall still works by matching words.

#Setup and switching

#I already use Hivemind. What happens?

honeycomb and Hivemind are siblings and share one sign-in, but running both at once isn't supported. When you set up honeycomb, the hive dashboard notices an existing Hivemind install and offers to move you over cleanly, usually without needing to sign in again.

#Can I use honeycomb across multiple machines?

Yes. Sign in on each machine with the same account, and a memory captured on one is available on the others.

#How do I see what honeycomb knows?

Open the hive dashboard at 127.0.0.1:3853 (or run honeycomb dashboard). It shows your memories, your connected assistants, your shared skills, a map of your codebase, and overall health.

#How do I remove honeycomb?

honeycomb uninstall reverses only the changes honeycomb made to your tools, leaving everything else untouched.

#Can I run honeycomb against my own storage instead of the hosted one?

Yes. honeycomb's storage layer is pluggable, and the daemon is the only process that talks to it, so pointing at a self-hosted backend (Activeloop's open-source pg_deeplake Postgres extension) is a single decision made at login with honeycomb login --endpoint <url>.

#Is honeycomb open and free to use?

Yes, licensed under AGPL-3.0-or-later. You can use it commercially or privately at no charge, as long as you keep the license notices and share your source if you run a modified version as a network service.

#Common questions

What is honeycomb in one sentence? A shared, lasting memory for your AI coding assistants, so what one of them learns is remembered by all of them, across sessions, tools, devices, and your team if you want.

Where does my data live? In your own Deeplake store, which you control and can host in your own cloud account. The local daemon is the only thing that connects to it.

Does honeycomb cost money in AI model usage? Everyday capturing, recalling, and priming don't need their own AI model or API key. Two optional extras, turning sessions into skills and the self-tidying loop, can use a model, and both are opt-in.

I already use Hivemind. What happens? honeycomb and Hivemind share one sign-in, but running both at once isn't supported. The hive dashboard detects an existing Hivemind install and offers to move you over cleanly, usually without needing to sign in again.