Skip to content
allman

Introduction

allman syncs your LinkedIn conversations to your machine and gives you three ways to work with them: a CLI, a terminal inbox, and a native Mac app. Read, search, reply, and pipe messages to AI agents — all from one local store.

What is allman?

Section titled “What is allman?”

allman is one CLI with two front-ends on top of it:

  • allman CLI — the base. It owns the local store, the sync, the rate limits and every call to LinkedIn. Sync, search, send, listen and manage your messages from the command line; every command supports --json for piping to scripts and AI agents.
  • allman-tui — a full terminal inbox with vim-style navigation, live updates, emoji reactions, and keyboard-driven workflows. Bundles the CLI.
  • Allman for Mac — a native macOS menu bar app: an iMessage-style inbox with labels, notes, snooze, group messages, your network and insights. Keeps the store live in the background and sends native notifications. Bundles the CLI.

Both front-ends read the store directly and shell out to the CLI for every write, so there is exactly one code path that talks to LinkedIn.

Design principles

Section titled “Design principles”

Local-first

Section titled “Local-first”

Messages are stored as plain JSONL files in a local git repository. Nothing is sent to any server. Your data stays on your machine.

AI-native

Section titled “AI-native”

NDJSON event streams, structured JSON output, and a scriptable CLI. Built for agents that manage your conversations.

Keyboard-driven

Section titled “Keyboard-driven”

Full terminal UI with vim bindings, inline search, emoji reactions, and everything you need without leaving the terminal.

How it works

Section titled “How it works”
  1. Authenticate via a real browser session (Playwright Chromium). Your cookies stay local.
  2. Sync pulls your conversation history into local JSONL files.
  3. Read, search, reply from the CLI, the TUI or the Mac app. All reads come straight from disk — fast and offline-capable.
  4. Listen streams real-time events (messages, reactions, typing indicators) as NDJSON.
  5. Send replies with pre-send safety checks and built-in rate limiting.

Next steps

Section titled “Next steps”