/docs/

Talkie CLI

Access your voice memos, dictations, and workflows from the command line. Designed for agents and power users who want programmatic access to their voice data.

Installation

The CLI is available as an npm package. It reads directly from Talkie's local SQLite database — no server required.

Install via npm

bun add -g @talkie/app

# Or CLI only
bun add -g @talkie/cli

# Or with the one-liner (installs CLI + app)
curl -fsSL go.usetalkie.com/install | bash

Once installed, the talkie command is available globally. All commands output JSON by default when piped, and human-readable tables in the terminal.

Agent Access Answers

Can agents read Talkie data?

Yes. Agents can call the Talkie CLI to read memos, dictations, searches, workflow runs, transcription results, and local inference output as JSON.

Does the CLI need a server?

No. The CLI reads from the local Talkie library and works without a hosted API for normal memo, dictation, search, and workflow history access.

What tools compose with it?

The output is designed for jq, shell scripts, Claude Code, Codex, MCP-compatible agents, and local model CLIs.

Global flags (all commands)

--json Force JSON output|--pretty Force table output|--since Filter by date (7d, 24h, 2025-02-01)|--limit Max results

Commands

talkie memos

List and view voice memos with transcripts, summaries, and tasks

$talkie memos

$talkie memos --since 7d --limit 10

$talkie memos a1b2c3 # Get memo by ID prefix

$talkie memos a1b2c3 --segments # With word timestamps

--since--limit--segments

talkie dictations

List keyboard dictations with app context metadata

$talkie dictations

$talkie dictations --since 24h

$talkie dictations f9e8d7 # By ID prefix

--since--limit

talkie search

Full-text search across all recordings with powerful filters

$talkie search "product launch"

$talkie search "meeting" --type memo

$talkie search "notes" --app "Chrome"

$talkie search "recent" --sort newest --has summary

$talkie search "long" --longer-than 60

--type--app--source--has--sort--longer-than--shorter-than--since--until

talkie workflows

View workflow execution history with step-by-step details

$talkie workflows

$talkie workflows --status completed

$talkie workflows --status failed --since 7d

$talkie workflows a1b2c3 # Full run details

--status--since--limit

talkie stats

App statistics — dictation counts, streaks, top apps

$talkie stats

$talkie stats | jq .streak

talkie engine

Manage the transcription engine — status, models, transcribe files

$talkie engine status

$talkie engine models

$talkie engine transcribe ~/audio.m4a

$talkie engine transcribe ~/audio.m4a --model parakeet:v3 --timings

--model--timings--raw

talkie inference

Local LLM inference — generate text, chat, manage models

$talkie inference status

$talkie inference models

$talkie inference generate "Summarize this meeting"

$talkie inference chat --model llama:7b

--model--system--temp--tokens--verbose

For Agents

The CLI is designed for AI agent integration. All commands output structured JSON when piped, making it easy to compose with tools like jq, Claude Code, or any MCP-compatible agent.

Agent examples

# Feed recent memos to an agent
talkie memos --since 7d | jq '.[].text'

# Search for context on a topic
talkie search "project alpha" --type memo | jq '.[0]'

# Check workflow failures
talkie workflows --status failed | jq '.[] | {name: .workflowName, error: .errorMessage}'

# Transcribe and process a file
talkie engine transcribe recording.m4a | jq '.text'

Every command supports prefix ID matching — you don't need the full UUID, just enough characters to be unique. This makes it natural for agents to reference specific recordings.

PreviousWorkflows

NextExtensibility