Troubleshooting

The apiary is built to degrade quietly rather than crash, so the most common "problem" is something running in a reduced mode without telling you plainly. Start with the dashboard, it names any degraded subsystem in plain language before you have to guess.

#Where should I look first?

Open the dashboard:

honeycomb dashboard

The health strip names any subsystem that is down, storage, the semantic-search model, or a missing piece of the store. The recall bar shows a "lexical fallback" badge when the last recall ran by matching words rather than meaning. Most of the symptoms below can be confirmed here at a glance.

#The command is not found right after install, what do I do?

Installing a global command does not refresh the current terminal's path, the command exists but this shell does not know about it yet. Open a new terminal window and run honeycomb status again, a fresh shell picks it up. If it is still missing in a new window, reinstall with the one-command installer, which resolves the command's full location during setup.

#The installer stopped and printed a command to run myself, why?

A step needed elevated permission it could not get, commonly installing Node on a locked-down machine. This is by design, the installer stops cleanly rather than leaving a half-finished state. Run the exact command it printed, then paste the original install line again. It is safe to re-run and picks up where it left off.

#The dashboard did not open on its own, what do I do?

The opener could not launch your browser, which is non-fatal. Copy the local dashboard address from the installer output, or run:

honeycomb dashboard

If the page still does not load, the daemon may not be running. See the next question.

#The daemon is not reachable, what do I do?

honeycomb status reports the daemon as down, or dashboard pages do not load. Run a command that nudges it up:

honeycomb status
honeycomb setup

The daemon is safe to start repeatedly, an already-running one is left alone. If status keeps reporting it as unreachable, restart your machine to clear any stuck process, then try honeycomb status again.

#Recall comes back empty for something I know I saved, why?

Check these three things, in order.

You are in a different project folder. Memory is scoped to the project folder you are working in, so a note saved in one repository does not surface while you work in another. Confirm your current project with honeycomb project list.

Semantic search has not warmed up yet. Until the small local model is ready, recall matches words rather than meaning, so a note phrased differently from your query might not surface. Open the dashboard and check the recall bar for a "lexical fallback" badge. If you see it, rephrase closer to the note's original wording, or wait for the model to finish warming up.

The note was never written. If you were in read-only mode for that session, nothing new got saved. Confirm you are not in read-only mode, then save the note again.

#Recall results feel thin or worse than usual, why?

Recall is likely running in its word-matching fallback because semantic search is off or a part of the store is temporarily unavailable, and the system keeps working quietly so this is easy to miss. Open the dashboard, the health strip names any down subsystem and the recall bar shows the lexical fallback badge when meaning-search was not used.

#An assistant is not being wired by setup, why?

Setup only wires an assistant it can detect on your machine, meaning that assistant's own config folder exists where the apiary looks. If the assistant is installed somewhere non-standard, or not actually installed, it gets skipped. Confirm the assistant is installed, then wire it directly by name:

honeycomb connect <assistant>
honeycomb status

Supported names today are claude-code, cursor, and codex, with hermes, pi, and openclaw in progress.

#An assistant is wired but does not seem to recall, why?

Two likely causes. The daemon may have been down at session start, in which case the briefing was skipped and your assistant kept working normally, confirm with honeycomb status and start a fresh session. Or you may be in a brand-new or different project that has nothing to recall yet, confirm with honeycomb project list and save a note or two to seed it.

#I switched company or team and queries look wrong, why?

Your active company changed but your sign-in was still bound to the previous one. The apiary detects this drift and re-mints a corrected token automatically at the start of a session. Start a new session, and confirm where you landed:

honeycomb org list
honeycomb workspace list
honeycomb project list

If queries stay wrong after a fresh session, switch explicitly with honeycomb org switch <name> and honeycomb workspace use <name>.

#I already had Hivemind and setup warned me, what do I do?

Running honeycomb and Hivemind together is not supported, and the dashboard detected a prior Hivemind install. Choose "proceed with honeycomb" in the dashboard, it backs up your Hivemind setup first, then moves you over. Because the two share one sign-in, you usually will not even need to sign in again.

#Common questions

#The command is not found right after install, what do I do?

Open a new terminal window and run honeycomb status again. Installing a global command does not refresh a terminal that was already open, a fresh shell picks up the new command.

#The daemon is not reachable, what do I do?

Run honeycomb status, which probes the daemon and starts it if it is absent, or run honeycomb setup to bring it up. If it still will not come up, restart your machine to clear any stuck process and try again.

#Recall comes back empty for something I know I saved, why?

Check three things in order: you are in the same project folder where you saved the note, semantic search has warmed up (the dashboard shows a lexical fallback badge if not), and you were not in read-only mode when you saved it.

#An assistant is not being wired by setup, why?

Setup only wires an assistant it can detect on your machine. Confirm the assistant is actually installed, then wire it directly with honeycomb connect followed by its name, and check with honeycomb status.

#I switched company or team and queries look wrong, why?

Your sign-in was still bound to the previous org. Starting a new session triggers an automatic correction. Confirm where you are with honeycomb org list and honeycomb workspace list.