Glossary

These are the words you will see across every page of the apiary docs. Each one gets a short, plain answer. If a term is specific to one product, its own docs go into more depth.

#What is a daemon?

A daemon is a small program that runs quietly in the background on your machine. In the apiary, each product (honeycomb, nectar, doctor, hive) runs its own daemon, and the daemon is the only thing that talks to your memory store. That keeps storage, security, and consistency in one place instead of scattered across every tool you use. You rarely start or stop one by hand, it comes up when needed.

#What is a harness?

Harness is the technical word for the AI coding tool you actually type into: Claude Code, Cursor, Codex, and (in progress) Hermes, pi, and OpenClaw. "Agent" and "assistant" mean the same thing. The apiary plugs in underneath the harness so capture and recall happen automatically as you work in it.

#What is capture?

Capture is the quiet recording of what happens while you work: your prompts, the actions your assistant takes, and the results. It is cheap, instant, and never blocks your assistant. If capture ever hiccups, your assistant just keeps working normally. Capture is raw material, it gets turned into clean notes later.

#What is recall?

Recall is asking for the right memory back at the right moment. It runs automatically at the start of a session as a short briefing, and your assistant can ask for more any time during the session. Recall matches both the exact words you use and, once warmed up, the meaning behind them.

#What is a memory?

A memory is a clean, distilled note kept about your work, a decision, a fix, a convention, or a gotcha you would otherwise re-explain every morning. Memories are what recall hands back.

#What are the three tiers (key, summary, raw)?

The three tiers are the same memory stored at three levels of detail. The key is a one-line headline. The summary is a short recap. The raw is the full original record. You skim keys, open a summary if it looks useful, and only read the raw detail when you genuinely need it, the same way a person remembers a gist first and fills in details on demand.

#What is priming?

Priming, or the prime, is the small "here is what I already know about this project" briefing handed to your assistant at the start of a session. It is deliberately short, just headlines, so it informs without cluttering the conversation.

#What is a skill?

A skill is a reusable lesson, written once and shared. When you or a teammate solve something worth keeping, like a migration trick or a debugging routine, the apiary can turn it into a skill that shows up automatically for everyone on the team, no copying files around.

#What is semantic search?

Semantic search is the optional ability to find a memory or a piece of code by what it means, not just by matching exact words. Turn it on and you catch the "I didn't know the exact term to search for" cases. Turn it off and recall still works by matching words, it just finds fewer of those cases.

#What are org, workspace, and project?

These are the three nested levels that keep memory in the right lane on a team. Org is your company, the outer boundary, two companies never see each other's anything. Workspace is a team within that company, kept separate from other teams. Project is the specific repository or folder you are actually working in, so a note from one repo does not surface while you work in another. You do not manage the project level by hand, the apiary figures it out from the folder you are in.

#What is Deeplake?

Deeplake is the database for AI, built by Activeloop, where the apiary's memories and codebase understanding live. It is good at exact lookups and meaning-based search at the same time, it keeps a full version history, and it can run in your own cloud account. Always one word, capital D.

#What is Hivemind?

Hivemind is Activeloop's open-source agent-memory project. Honeycomb, the memory product in the apiary, is built on top of it, and the two share a sign-in.

#What is a knowledge graph?

A knowledge graph is honeycomb's map of the things in your work, people, projects, tools, and decisions, and how they connect. It is what lets memory answer "what is true right now, and what does it depend on," not just "what did I say once."

#What is a codebase graph?

A codebase graph is a map of your actual code: its files, functions, and how they call and import each other. It lets an assistant answer questions like "what would changing this break" grounded in your real project instead of a guess.

#What is MCP?

MCP, the Model Context Protocol, is a standard way for AI tools to call external helpers. The apiary runs an MCP server so any assistant that speaks MCP can ask for memory or code understanding directly, as a native tool in its own list, instead of you gluing the two together yourself.

#What is ROI in the context of the apiary?

ROI is the time you get back. It shows up as fewer re-explanations of your project to a fresh assistant session, fewer times someone re-solves a bug that was already fixed, and less copy-pasting the same convention across tools and teammates. The apiary is built to compound that over time instead of losing ground as memory grows.

#Common questions

#What is a daemon?

A daemon is a small program that runs quietly in the background on your machine. In the apiary, daemons are the only things that talk to your memory store, so everything stays in one consistent, safe place. You rarely deal with one directly.

#What is a harness?

Harness is the technical word for the AI coding tool you actually use, like Claude Code, Cursor, or Codex. It is the thing the apiary plugs underneath so it can capture and recall automatically.

#What is capture?

Capture is the act of quietly recording what happens as you work: your prompts, what the assistant did, and what came back. It is the raw material honeycomb later turns into clean notes.

#What is recall?

Recall is asking for the right memory back at the right moment. It happens automatically at the start of a session and again whenever your assistant needs more detail.

#What is a memory?

A memory is a clean, distilled note about your work, a decision, a fix, a convention, or a gotcha. Memories are what recall returns.

#What are the three tiers?

The three tiers are the same memory kept at three levels of detail: a one-line key, a short summary, and the full raw record. You skim the key, open the summary if it looks relevant, and read the raw detail only when you truly need it.

#What is priming?

Priming is the short briefing handed to your assistant at the start of a session, so it starts already knowing your project's recent decisions instead of starting blank.

#What is a skill?

A skill is a reusable lesson, written once and shared automatically with you and your teammates, so a trick one person learns does not have to be rediscovered by everyone else.

#What is semantic search?

Semantic search finds memories or code by meaning rather than exact wording, so you can find the right note even if you would not have guessed its exact phrasing.

#What are org, workspace, and project?

Org is your company, workspace is a team within it, and project is the specific repository or folder you are working in. The apiary keeps memory separated across all three so the right people see the right notes.

#What is Deeplake?

Deeplake is the database for AI, built by Activeloop, where the apiary's memories and codebase data are stored. It handles both exact lookups and meaning-based search, keeps full version history, and can run in your own cloud.

#What is Hivemind?

Hivemind is Activeloop's open-source agent-memory project. Honeycomb, the memory product in the apiary, is built on top of it.

#What is MCP?

MCP, the Model Context Protocol, is a standard way for AI tools to call external helpers. The apiary offers an MCP server so assistants that speak MCP can ask for memory directly as a built-in tool.

#What is ROI in the context of the apiary?

ROI here means the time you get back from not re-explaining your project, not re-solving problems you already solved, and not repeating the same fix across every tool and every teammate.