Using AI puts you at one of six levels: one-shot, collaborator, reviewer, approver, audience, farmer. Each step up increases efficiency at the price of reduced understanding and increased token costs. Most AI products choose the level for you — so you end up using six or more of them.
With toast you can roast your directory, debug your laptop, analyze logs. You can build one-line email and imessage bots. You can use cron to create a daily newsletter that summarizes the latest research in your field. toast may just be the best way to learn AI. Feel free to talk with toast itself — see the live chat bubble in the lower-right corner of this page.
One tool, six levels of AI. We build toast so you can choose the right level for the problem you are solving — Unix-native AI: pipes, dotfiles, exit codes. Below, the teal lines are your keystrokes — watch them thin out as the levels climb.
One command in, one answer out.
$ ls -al | toast "roast my directory" files-1 through files-213? That's not a directory structure, that's a cry for help. You have `toasted-15.cpp` and `v-9.c`. Git exists. Use it. `Claude.dmg`, `Ollama.dmg`, `Second_Life`... stop hoarding installers from 2024. `thephonebetweenus` has 5 versions. Make a decision. 16GB LLM file sitting next to a `.DS_Store`. Respectable trash heap. $ ps aux | toast "anything suspicious" Found a few things: 1. **VidyoConnect**: Running from a temp folder (`AppTranslocation`) with `--no-sandbox`. Drag it to `/Applications` and reinstall to fix security flags. 2. **Onnai.app**: Located in Xcode `DerivedData`. Is this your debug build? 3. **Zombies**: 2 defunct processes (PIDs 44310, 44023) need cleanup. 4. **AI Load**: LM Studio and Ollama are active and using resources. $ history | toast "what was I trying to do yesterday" Moved files, opened the San Mateo call list, and checked your site's HTML. Then you roasted your directories and investigated high CPU usage.
One question, one answer, back to your prompt. Pipe anything in —
logs, configs, ps output — and project context rides along from the
nearest .crumbs up the tree. Every command is yours.
You drive; toast thinks alongside.
A bare toast opens chat — same tool, but now context carries
across turns. Ask, follow up, drill in:
$ toast --- > @models.py explain this Three SQLAlchemy models — User, Session, Token. User owns Sessions; the cascade on delete is doing more than you probably intend. > which cascade? Session.tokens. Deleting a user nukes their sessions and every token — including the ones you keep for audit. Split the relationship. > /exit
History lives in ./.chat — leave and come back, and the
conversation doesn't replay, it continues. That makes chat the ideal test
bench for a .persona: drop the file in the directory, talk to it, tune
it. @file inlines any file into context. Every next move is yours,
which is why you still understand the answer.
toast writes; you review it before it ships.
In Unix, the reviewer can be another toast — stdout is stdin, so adversarial review is a one-liner. Personas are just symlinks to the binary:
$ toast --add Dev && toast --add Paranoid Added Mimic: /usr/local/bin/Dev -> /usr/local/bin/toast Added Mimic: /usr/local/bin/Paranoid -> /usr/local/bin/toast $ cat auth.py | Dev "fix the token refresh" | Paranoid "what's wrong with this patch" The patch works but opens a race — two concurrent refreshes both pass the expiry check. And the old token stays valid 30s after rotation.
Piped toast is read-only — its verdict lands in the next pipe, not on
disk. Prefer a live argument? toast --room dev,paranoid puts both
personas in one chat and lets you referee. Run them on toasted and the code never
leaves your machine.
You stop reading the work and start reading signals about it.
Hand toast a file and it patches the file itself — atomically, at scale:
$ find . -name "*.py" -exec toast {} "add type hints" \; [toast] Updated api.py [toast] Updated models.py (warning: 2 matches found, replaced first only) ... 197 more ... $ mypy . && ito log "added type hints" Success: no issues found in 199 source files # files patched: 199 · files you read: 2
A green check tells you the code passes the check — not that the check
covers what matters. Did you read the warning on models.py, or did you
read Success? Level 4 is a fine place to work, as long as you know that's
the trade.
Nobody's at the prompt.
Autonomy is opt-in: .tools is an allowlist you write yourself, one
command per line — no file, no tool calls. Put toast on the list and
the agent's favorite tool is another toast. Cron for the routine, loops for the
open-ended — toast exits 1 when the model prints DONE,
so the shell itself decides when work is finished:
# the allowlist — one command per line. note who's on it $ printf 'df\njournalctl\nsystemctl\ntoast\n' > .tools # daily briefing 0 7 * * * curl -s https://news.ycombinator.com | sed 's/<[^>]*>//g' | toast "top 5 articles by novelty" | mail -s briefing you@example.com # monday, 9am — you are now absent from your own standup 0 9 * * 1 ito history | toast "write my standup, first person, modest" | mail -s standup you@example.com # loop until the model itself says stop while toast server.py "refactor until clean; print DONE when finished"; do :; done # weeks later, something feels off $ ito history | toast "what was the focus last week"
One teal line left — and it's you asking the AI to narrate your own
machines back to you. Every tool call runs through
jam, five rounds max — and with
toast in .tools, toast can call toast: delegation all the
way down. This is where
UnixClaw lives — use it on
purpose, not by drift.
You plant; the system grows it; you harvest.
The audience automates tasks — a farmer grows deliverables. Seed a
directory with an outline, a .persona, and a .tools that
includes toast itself, set the loops running, and come back for the harvest: a
manuscript, a go-to-market plan, a codebase:
# plant: an outline, a voice, and tools — toast included $ mkdir novel && cd novel && vim outline.md .persona $ printf 'toast\nEditor\n' > .tools # grow: chapters draft themselves, then twenty polish passes (jam) 12 times toast outline.md "draft the next chapter into its own file; keep continuity notes in .crumbs" 20 while Editor manuscript.md "polish for publishing" # another field: the go-to-market plan, every monday night 0 22 * * 1 toast brief.md "grow the GTM plan — positioning, pricing, launch, competitor scan via toast; DONE when complete" # harvest $ wc -w ch*.md 84312 total # the notes the system kept for itself $ cat .crumbs Mara learns about the ledger in ch7 — don't re-reveal it later. The fire is day 3. Osei's limp: left leg. Chapters 9–12 are winter. $ cat launch-plan.md | toast "poke holes in this"
The crop is plain files — and so is the memory: the model writes
notes to .crumbs, and every later run in that directory reads them
back. wc the chapters, grep the diary,
ito log the milestones, pipe anything back through level 1 to
taste the harvest. This
is what
Gradient
Descent for Anything was built for: autonomous refinement that runs until the
model itself says DONE.
So — which level are you? The honest answer should be: it depends on the task. Using AI at the audience level and above has a cost: your mental model thins, and it gets surprisingly hard to spot a flawed conclusion you didn't participate in reaching. Efficiency may have improved, but your skill atrophies the more you delegate.
The toast solution is autonomy that stays interruptible —
pipes you can inspect, diffs you can review, exit codes you can gate on, an
allowlist you wrote yourself. Every other AI product sells only the top of the ladder. toast
sells all six — one tool, six levels — because the right question was never
how autonomous can it get but
at which prompt does your judgment earn its keep.
Everything above is one small C binary. Pipe text in, get Unix knowledge out — toast, jam, ito, toasted: simple tools for humans and AI alike, built for the people who keep the lights on.
Free on MacBooks. The installer sets up appled —
Apple Intelligence as your inference provider. No account, no key, fully on-device.
Outgrow it? pkill appled and toast offers $20 Pay and Go.
We see anonymized usage only, never prompts.
BYOK · free OpenAI, Anthropic, Google, Mistral, Groq, Cerebras, Perplexity, xAI, OpenRouter, Together
Local · free appled, toasted, Ollama, MLX, LM Studio, llama.cpp, vLLM, LocalAI, Jan
Levels 1–4 need nothing but toast. Levels 5–6 done right are machinery — an AI-native shell, version control that records intent, local inference, bots. That's the Member plan.
No quoting, no expansion, no $ surprises. Built-in loops, RPN
math, UDP multicast for multi-machine coordination.
Record intent, derive diffs. No staging area, no detached HEAD. Single C file, ~1,100 lines. Search by why, not what.
ito log "refactored auth to use JWT" ito history | toast "what was the focus last week" Learn more →From-scratch daemon for Apple Silicon. Qwen3-Next-Coder at ~100 tok/s, 0.6s to first word, C++ against MLX. Code never leaves the machine.
toasted start cat auth.py | Security "audit this" Learn more →Personality lives in your .persona file. One line to deploy,
and there's a Telegram bridge: toast --telegram.
Unix is deterministic. AI is not. The hard part is the boundary work of making them compose — that's what a membership funds.
.personaLightweight toast talks to a local toastd, which keeps
an HTTP/2 connection pool to linuxtoaster.com. Written in C to minimize latency.
With BYOK, toastd connects directly to your provider — your traffic never touches
our servers.
On MacBooks, the installer downloads appled — a local inference
provider using Apple Intelligence. Completely free: no account, no key, no cost
per token; prompts never leave your machine. Outgrow it? pkill appled
and toast brings up a Stripe page for $20 PayGo with more powerful models.
Got a PROVIDER_API_KEY set for Anthropic, Cerebras, Google Gemini,
Groq, OpenAI, OpenRouter, Together, Mistral, Perplexity, or xAI? Use
toast -p provider. Zero config, zero cost from us.
Yes — appled, toasted, Ollama, MLX, LM Studio, KoboldCpp, llama.cpp, vLLM, LocalAI, or Jan. No internet, no API keys, full privacy.
A shell rebuilt for AI. No quoting, no expansion, no $ syntax —
strings just work; unrecognized input goes to the AI. Includes
set/get, while/times loops,
RPN math, and a UDP multicast basket for multi-machine coordination.
A from-scratch local inference daemon for Apple Silicon (Member tier). Loads Qwen3-Next-Coder — a 30B coding model — via C++ against Apple's MLX API. ~100 tok/s generation, ~400 tok/s prefill, 0.6s to first token. 128 GB supports 8/6/4-bit quantization, 64 GB supports 4-bit.
Locally. Context in .crumbs, conversations in .chat,
tool permissions in .tools. Version them, grep them, delete them.
Companies funding the rewrite of Unix. Your team gets a software license, priority support, and consulting options — and you're funding tools that make software simpler for all LinuxToaster users. Talk to us.
macOS and Linux today.
Free on MacBooks with appled — no account needed. $20 PayGo gets you a membership plus $20 in AI credits; top off anytime. Inference is charged by use; BYOK and local inference are free. We collect anonymized usage (model, token count) — never your prompts. Consulting and FDE are optional add-ons.
