Nectar reference
Every nectar command and the daemon's HTTP endpoint, with a plain description of what each one does.
Nectar is used through the nectar command-line tool and a daemon running on 127.0.0.1:3854. Core commands cover the daemon service, scanning and describing files with brood, and searching with search.
Nectar has one command-line tool and one daemon. This page lists what each command does. There's nothing here you need to configure to get started; see getting started with nectar for the normal path.
#What runs the daemon?
| Command | What it does |
|---|---|
nectar daemon |
Starts the daemon on 127.0.0.1:3854, serving a /health endpoint. |
nectar install |
Registers the daemon as an OS service, supervised by doctor. |
nectar uninstall |
Removes the OS service registration. |
nectar service-status |
Reports whether the OS service is currently running. |
#What scans and describes my files?
| Command | What it does |
|---|---|
nectar brood --dry-run |
Previews the full scan's cost, locally. Makes no model calls and writes nothing. |
nectar brood |
Reads and describes every file once. |
nectar brood --force |
Writes a fresh description for every file, ignoring anything already stored. |
nectar brood --limit N |
Describes at most N files at a time, useful for pacing cost. |
#What does searching?
| Command | What it does |
|---|---|
nectar search "<query>" |
Searches described files by meaning. |
nectar search "<query>" --limit N |
Limits the number of results returned. |
nectar search "<query>" --json |
Returns results as machine-readable JSON. |
#What else is there?
| Command | What it does |
|---|---|
nectar rebuild-projection |
Regenerates the committed .honeycomb/nectars.json file from Deeplake. |
nectar review-matches |
Shows low-confidence file matches for you to confirm before they're accepted. |
nectar prune --confirm |
Removes identities for files that have been missing for a long time. |
nectar --help |
Shows the full command list. |
#What does nectar need to actually scan and describe files?
A real scan that writes new descriptions needs two things already set up by the rest of the Apiary: credentials for the description model, and the description gateway enabled. If either is missing, the daemon still starts and answers /health, but scanning stays paused and tells you why. A dry run and search don't need either of these.
#Is there telemetry?
Anonymous, aggregate usage telemetry is on by default. It never includes file contents or file paths. Turn it off with NECTAR_TELEMETRY=0, or with the standard DO_NOT_TRACK variable if you already use that.
#Are there other ways to reach nectar besides the CLI?
Yes. A few interfaces exist beyond the command line:
- The daemon exposes endpoints under
/api/hive-graph/*on127.0.0.1, which the hive portal reads. - A typed SDK ships as
@legioncodeinc/nectarfor anyone building on top of it directly. - Nectar doesn't run its own MCP server. Its results reach your assistant through honeycomb's recall tools instead.
#Common questions
How do I preview the cost of a scan before running it?
Run nectar brood --dry-run. It makes no model calls and writes nothing.
How do I search my codebase by meaning?
Run nectar search "<your question>", or just ask your assistant directly.
Does nectar have its own MCP tools? No. It has no MCP server of its own. Its results surface through honeycomb's recall tools.