CLI

kura <command> [options]. Run kura --help for the command list and kura <command> --help for a command’s flags.

Global conventions

--json — every read command can emit machine-readable JSON instead of formatted text. Handy for scripts and consumed by the MCP server.
Document references — commands that take a document accept a doc key, a #key, a full path (db/sqlite/Notes), or a unique title (e.g. kura get "Today's note" or kura get #a1b2c3).
NO_COLOR — set it to disable colored output.
Data location — KURA_HOME overrides the data directory (default ~/.kura); KURA_DB overrides the database file.

Command	What it does
`kura init`	Create `~/.kura`, download the tokenizer, create the DB
`kura doctor`	Diagnose SQLite, extensions, LLM providers, DB integrity
`kura doctor --fix`	Repair the index, GC orphans, re-resolve links, detect embedding-model changes
`kura status`	Store statistics: counts, embedding coverage, stale candidates
`kura config list\|get\|set`	Read and write `~/.kura/config.toml`

Command	What it does
`kura add <file>`	Add a document; `-` reads stdin; `--path`, `--tags`, `--title`, `--bucket`
`kura get <ref>`	Print a document (Markdown or `--json`)
`kura edit <ref>`	Open in `$EDITOR`; frontmatter edits title, path, and tags too
`kura rm <ref>`	Delete a document
`kura mv <ref> [<title>]`	Rename and/or move (`--path`); rewrites wiki links. `--prefix <old> <new>` moves a whole subtree
`kura mv suggest`	Propose paths for unfiled documents from links/tags/search signals (plus the LLM when available); `--apply`, `--json`
`kura ls`	List documents; `--tag`, `--bucket`, `--prefix`, `--sort`, `--stale`
`kura clip <url>`	Capture a web page, cleaned up by the LLM; filed under `clip.path` (default `clips`)
`kura export --dir <d>`	Write documents as Markdown with frontmatter; paths become subdirectories
`kura import <dir>`	Round-trip via `kura_key` (update or create); subdirectories become document paths

Command	Mode	Notes
`kura search <q>`	Keyword	FTS5 BM25; always available; < 100 ms
`kura vsearch <q>`	Semantic	Vector KNN over local embeddings
`kura query <q>`	Hybrid	Fuse keyword + vector, then local-LLM rerank
`kura embed`	—	(Re)generate embeddings for documents missing them

See Search for how the modes differ and their degraded behavior without an LLM provider.

Command	What it does
`kura bucket ls\|add\|rm`	Manage buckets
`kura tag ls`	List tags; `--tree` for the hierarchy
`kura tag suggest`	LLM tag suggestions; `--untagged`, `--apply` (reuses your taxonomy)
`kura tag audit`	Merge candidates for similar tags; oversized-tag warnings
`kura link ls <ref>`	Outlinks, backlinks, and 2-hop links for a document
`kura link broken`	List unresolved wiki links

Command	What it does
`kura browser`	Serve the browser UI at `http://127.0.0.1:7578`
`kura mcp`	Run the MCP server on stdio for AI agents
`kura mcp --print-config`	Print an `.mcp.json` snippet

Code	Meaning
`0`	Success
`1`	Error (including conflicts, e.g. a unique-constraint violation)
`2`	Usage error (bad arguments)
`3`	Not found (no such document)
`4`	LLM provider unavailable

Scripts can rely on these: exit 4 specifically means “the feature needs a model and none was reachable”, distinct from a genuine error.