> ## Documentation Index
> Fetch the complete documentation index at: https://docs.omi.me/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> Talk to Omi from your terminal. Designed for humans and agents alike.

`omi-cli` is the command-line interface to the Omi developer API. It exposes
scoped, JSON-friendly verbs for the four primary nouns Omi maintains about you,
so you can drive Omi from shell pipelines, CI jobs, AI agent harnesses, or your
own personal scripts.

* **PyPI:** [pypi.org/project/omi-cli](https://pypi.org/project/omi-cli/)
* **Source:** [github.com/BasedHardware/omi/tree/main/sdks/python-cli](https://github.com/BasedHardware/omi/tree/main/sdks/python-cli)
* **Install:** `pipx install omi-cli`

<CardGroup cols={2}>
  <Card title="Memories" icon="brain">
    Facts and learnings the system knows about you.
  </Card>

  <Card title="Conversations" icon="comments">
    Captured & processed audio/text exchanges.
  </Card>

  <Card title="Action items" icon="list-check">
    Tasks and follow-ups extracted from your captures.
  </Card>

  <Card title="Goals" icon="bullseye">
    Tracked progress metrics over time.
  </Card>
</CardGroup>

***

## Why a CLI?

Omi already ships:

* a **Python SDK** ([`omi-sdk`](/doc/developer/sdk/python)) for Bluetooth and audio capture,
* a **Flutter mobile app**, **macOS desktop app**, and
* an **MCP server** ([`MCP`](/doc/developer/mcp/introduction)) for Claude Desktop and similar tools.

What was missing was a **terminal-native, scriptable, JSON-emitting** interface
to the data plane — the thing every modern dev tool ships (`gh`, `stripe`,
`vercel`, `linear`) and the thing AI agents reach for first.

`omi-cli` fills that gap. The same binary works equally well for a human
reading their inbox of action items at the terminal and for an LLM agent
piping `--json` output through `jq`.

## What you can do with it

```bash theme={null}
# Read your memories with a category filter, just the IDs and content as JSON:
omi --json memory list --categories work,skills | jq '.[] | {id, content}'

# Pipe the transcript of a meeting into a new conversation:
cat meeting_notes.md | omi conversation create --text - --text-source message

# Mark every action item older than 30 days as complete:
omi --json action-item list --open \
  | jq -r '.[] | select(.created_at < (now - 30*86400 | todate)) | .id' \
  | xargs -I{} omi action-item complete {}

# Track progress on a daily goal in a cron job:
omi goal progress g_xyz "$(get_today_progress)"
```

## How it relates to MCP

The two are complementary, not competing:

* **MCP** is the right surface for AI assistants embedded in apps that speak
  the Model Context Protocol (Claude Desktop, Cursor). It's invisible to you —
  the LLM calls tools directly.
* **CLI** is the right surface when you want a stable command line your
  scripts, agents, and shell pipelines can drive — anywhere a process can run.

Both hit the same Omi API and respect the same scopes.

## What's next

<CardGroup cols={2}>
  <Card title="Install" icon="download" href="/doc/developer/cli/install">
    `pipx install omi-cli` and you're done.
  </Card>

  <Card title="Quickstart" icon="rocket" href="/doc/developer/cli/quickstart">
    Your first five commands, end to end.
  </Card>

  <Card title="Command reference" icon="book" href="/doc/developer/cli/commands">
    Every verb, every flag.
  </Card>

  <Card title="For agents" icon="robot" href="/doc/developer/cli/agents">
    Agent contract, exit codes, JSON shape.
  </Card>
</CardGroup>