Skip to main content
Five copy-pasteable scenarios that cover most everyday CLI workflows. Each runs in under 30 seconds against a populated tenant.

​
1. List your most recent notes

Pretty output in a TTY:
Add --json to switch to NDJSON for piping:

​
2. Search by keyword and date range

tiro notes search uses Tiro’s deep search — each result is a note hydrated with its primary documents (one-pager, custom). The keyword is required (positional or --keyword). Date filters accept ISO-8601 or relative shorthand:
Folder-scoped:

​
How keyword matching works

tiro notes search queries Tiro’s full-text search index over note titles and paragraph content. Matches are case-insensitive and tokenized — "OKR planning" returns notes containing both “OKR” and “planning”, not the literal phrase. Results are reordered by relevance with createdAt desc as tiebreaker. The deep variant (this command, POST /v1/external/notes/search) also hydrates each note with its primary documents (one-pager, custom) so an LLM client can read content alongside metadata in one call. The lighter tiro notes list --keyword "..." returns metadata only. Without --keyword, tiro notes list orders results by createdAt desc. With --keyword, pagination cursor is always null (the response is bounded by what the search index returns).
Placeholder filter. Both tiro notes list and tiro notes search filter out placeholder notes by default (title === "Untitled" or sourceType === "onboarding"). A page of N may return fewer than N visible notes — pass --include-untitled to surface them.

​
3. Save a single meeting note as Markdown

stdout returns one line of metadata — the actual content goes to disk:
The Markdown file contains a YAML frontmatter, the participant list, and a transcript with one elapsed-timestamp header per paragraph followed by speaker-attributed segments:
Pass --no-timestamps (v0.3.0+) to drop the ### mm:ss headers entirely — handy for raw saving and LLM ingest where the timestamps are noise:

​
4. Pipe search results to jq

--json always emits NDJSON for list/search responses — one note per line. Combine with jq -c (compact) for further pipelines, or head / tail to slice. If the response includes a next-page cursor, it’s emitted as a final NDJSON line keyed _cursor:
Pass that back as --cursor to continue:

​
5. Get a transcript with speaker labels

For agents that already understand MCP’s get_note_transcript payload, switch to --format json — the shape is byte-for-byte identical:

​
5. Explore the workspace wiki

The wiki is Tiro’s auto-extracted knowledge graph (entities, concepts, and their links) over your notes. These commands are read-only and require a wiki-enabled workspace (paid, opt-in).
--workspace is optional — omit it to use the default workspace for your API key. If your account belongs to more than one workspace, run tiro wiki workspaces first and pass the guid whose wiki is on. A gated workspace prints the backend’s upgrade message and exits non-zero. → Continue to For AI agents for tool-selection guidance and worked agent flows.