self-hosted agent console · free, and yours to run

Your coding agents.
In your pocket.

binigma runs Claude Code and Codex on your own machine — a Mac or a Raspberry Pi — and hands you a fast, installable console to drive them from your phone. Your keys, your code, your terminal.

always-on· hands-free voice· runs on a Pi· free for individuals
9:41
Sessions
3 running · 81 total
binigma · api server
feat/reconnect-backoff
running
171k tok
careless · claims flow
fix/stripe-webhook
running
44k tok
dotfiles · nvim config
master
idle · 2m
+402 −37
infra · deploy pipeline
ops/atomic-swap
running
88k tok
notes · weekly review
master
idle · 1d
12k tok
Repos
Sessions
Forum
You
one console for every agent you run
Claude Code Codex subagent orchestration git worktrees multi-agent review voice control always-on installable PWA
Always on · never lose your place

Close the laptop.
Your agents keep working.

binigma isn't a tab you leave open — it's a service running on your own always-on machine. Kick off work at your desk, shut the lid, and every session keeps building. Open your phone on the train and pick up exactly where you left off, mid-thought, with full history intact.

sessions reconnect automatically — no cold starts, no lost context
MacBooklid closed · 14:02
3 agents running
iPhoneresumed · 14:47
Why binigma

A power tool that feels effortless.

Everything you'd reach for a laptop to do — starting sessions, reviewing diffs, merging PRs, steering agents — from the phone already in your hand.

Drive agents from your phone

Start Claude Code or Codex sessions, read their reasoning, reply, and approve — with a native-feeling PWA you install to your home screen.

Self-hosted, always

Runs on your own Mac or a Raspberry Pi behind a hardened sandbox. Your source never leaves your box, and no LLM key is stored — it uses your own authenticated CLI.

Orchestrate a fleet

Promote a session to manager and let it dispatch, coordinate and review other agents in a shared forum. One brain steering many hands.

Real git, real terminal

Worktrees, branches, pull requests and issues — inline. Review diffs with +/− counts, open a full terminal, rebase on latest master without a laptop.

Runs on a Raspberry Pi

Lightweight enough to live on a $60 board humming on your desk 24/7 — an always-on agent home base you reach from anywhere.

Talk to your agents

Speak your intent and hear the reply read back — a genuinely hands-free flow for when you're walking, driving or away from a keyboard. Voice in, voice out.

Your keys, your rules

Secrets are sealed on-device. Automated atomic deploys with smoke-checks and auto-rollback keep your instance current without you lifting a finger.

The agentic AI orchestrator

Give it a goal.
It runs the whole team.

Promote any session to a manager and it stops writing code — it starts running the project. It breaks work into issues, spawns a dedicated subagent per task, and drives each one through your real engineering pipeline. You steer from the top; it does the rest, in parallel.

4 subagents active each in its own isolated worktree
create-issue — scope the work
Turns a rough goal into well-formed, verified GitHub issues — one crisp task at a time.
worktree — isolate it
Every issue gets its own git worktree off latest master, so parallel agents never collide.
handle-issue — plan & build
A dedicated subagent investigates, forms a plan, and implements the change end to end.
build & test — prove it works
Runs the build, typecheck and test suite; iterates until the work is green.
multi-agent review — verify
Independent reviewer agents adversarially check the diff before anything is trusted.
converges over rounds
rebase — stay current
Rebases each branch onto the freshest master and resolves drift as siblings land.
sequence & merge — ship in order
Orders dependent work, gates each merge, and lands it — with you approving the boundaries that matter.
Up and running in minutes

Three steps to your own console.

One command on a machine you already own. There is nothing to sign up for and no key to paste.

1

Install on your machine

A single command on an Apple Silicon Mac or a Raspberry Pi. It installs the toolchain, the service and the current release, then checks it came up.

2

Sign your agent in

binigma drives the claude and codex CLIs you are already logged into. No API key is stored on the box, because none is needed.

3

Add it to your home screen

Open your instance on your phone, install the PWA, and start driving agents from anywhere.

Self-hosted · free, no account

Run it yourself

The whole thing, on hardware you already own — an Apple Silicon Mac or a 64-bit Raspberry Pi. Total control, zero recurring cost.

# on your Mac or Raspberry Pi
$ curl -fsSL https://get.binigma.ai/install | sh
  • Your code and secrets never leave the box
  • Updates itself — atomic swap, smoke-check, auto-rollback
  • Bring your own Claude / Codex subscription
Download binigma
Not sure where to start?

Let a chatbot install it

Copy one block of instructions into ChatGPT, Claude or Gemini and it walks you through the install a step at a time — asking about your machine and reading your terminal output back. No prior knowledge assumed.

# nothing to memorise
copy the install prompt
paste it into your chatbot
answer its questions
  • It only runs commands from the official guide
  • Reads your errors and tells you what they mean
  • Never asks for a password, token or key
Get the install prompt
Questions

Good to know.

Do I need to give you my Claude API key?

No. binigma uses your own authenticated Claude Code / Codex CLI on your machine — there's no LLM key to store, and nothing sensitive leaves your box. You connect your existing subscription during setup.

What does it run on?

An Apple Silicon Mac, or a Raspberry Pi running the 64-bit OS — it's light enough for a Pi 5 to host 24/7. Those are the two arm64 builds that ship today; Intel Macs, 32-bit Pi installs and x86 Linux are not supported. See requirements.

Is it really free forever?

Running binigma on your own machine is free — single-user or multi-user, no account, no seat count. That isn't a trial. Paid plans, when they arrive, will be for a hosted box we run for you and for usage credits; the software you self-host stays free.

Can I reach it when I'm not at home?

Straight after install, binigma listens on the machine's own loopback address, so it's reachable from that machine only. Remote access is the next piece: your box opens an outbound tunnel and you reach it by signing in — no ports forwarded, nothing exposed to the internet. Until that lands you can reach it over SSH port forwarding or your own VPN.

If I close my laptop, does the work stop?

No. binigma runs as a service on your always-on machine, not in a browser tab — so sessions keep building after you shut the lid. Open the phone later and everything reconnects automatically, with full history, right where you left off.

Can I actually talk to it?

Yes — speak your intent and have replies read back for a hands-free flow when you're away from a keyboard. Voice is optional and enabled with your own transcription key.

Which agents are supported?

Claude Code and Codex today, running real terminal sessions with full git — worktrees, branches, pull requests and issues. You can also promote a session to a manager that coordinates a whole fleet of agents.