0.1.3Features

Everything Lunchbox does, on one sheet.

ARooms

Crew

The first agent in your crew answers you and brings in crewmates for the parts they do better.

Briefings

A crewmate gets the goal, what the room has learned, the files that changed and its own task.

Limits

When an agent runs out of usage, the next one takes the same task with a briefing and the first returns after its reset.

Proof

"Tests pass" counts only when a test command passed after the last edit. Without one, you decide.

Side by side

One message to the whole crew, answers in columns. Your pick feeds that agent's rating.

Search

Press ⌘K to search across all of Lunchbox from anywhere.

BControl

Access levels

Plan, Ask, Auto or Full access, chosen per message in the composer and enforced in the main process.

Blocked commands

A few commands such as rm -rf ~ stay blocked in every mode.

Checkpoints

The folder is saved before every message in a shadow history. Your own git history stays untouched.

Diffs and rewind

See what each message changed, rewind to any point, fork from it or replay it.

Exports

Save a room as Markdown or as an HTML replay that plays in any browser.

CPlans and cost

Pool

Plans and keys as one pool: what is left in each 5 hour and weekly window, when it resets, and ratings by kind of work.

Real limits

Plan limits are read through each tool's own login.

Savings

What prompt caching saved you, per agent and per room. Cached tokens cost about a tenth of the price.

How the saving is counted

Claude and OpenAI bill a token they have already cached at a tenth of the normal input price, so caching cuts up to 90% off those tokens. Lunchbox counts each cached token at that gap, minus what writing the cache cost. New input and output cost the same as before.

DWork without you

Flows

Scheduled runs with several steps. 25 templates in 5 groups, and each run gets its own room.

Library

55 playbooks, a catalog of MCP tools you add in one click, and packs.

Browser

A browser inside the app that agents can drive while you watch.

Phone remote

Follow and answer a room from your phone over Telegram.

EAgents

Command line agents

Claude Code, Codex, Grok Build and Gemini CLI, each with your own login and plan.

Connectors

OpenCode, Aider, Goose, Cursor agent, Qwen, Amp, Crush and Qoder.

Free models

OpenCode adds a dozen free models in one click, with no key. OpenRouter, Google AI Studio and GitHub Models have more with a free key.

Local models

Ollama, LM Studio and llama.cpp. Lunchbox can start Ollama and pull a model that fits your memory.

API keys

Anthropic, OpenAI, xAI, Google, OpenRouter, DeepSeek, Groq, Mistral, Together, Fireworks, Cerebras, Kimi, DeepInfra, Perplexity, NVIDIA and GitHub Models.

FYour data

No middle server

Requests go from your machine to the providers. There is no Lunchbox server in between.

Plain files

Rooms are .jsonl files on your disk.

Keychain

API keys are stored in the system keychain. Logins stay with each tool.

Updates

Lunchbox updates itself after the first install.