Right now, AI is a black box. You type a question, an answer pops out, and you have no clue what happened in between. So you trust it with small stuff and stay a little scared of the rest. And every word around it sounds designed to keep you out: agents, prompts, tokens, APIs.
Side opens the box. Say what you want in plain words: plan the birthday party, turn my recipes into a website, sort my inbox every morning. Side splits the job into steps on a board, like sticky notes, and gives each step to its own AI helper. You watch them work. You drag things around like in Notion. Anything important stops and waits for your yes. And around your third build, it clicks. Agent, memory, prompt. Not scary. You’ve watched them do their job.
Already deep into AI? Side is the control room you’re missing. Your models, keys, memory, agents and schedules on one canvas, across every machine you own. Start on easy mode, then go as deep as you want: routing, budgets, gates, a brain that learns you. Simple enough for your first agent. Deep enough for your hundredth.
opens on your desktop — localhost:4600, already yours
about 30 seconds · paste ANY model key — or explore the demo without one
Drop in the nodes yourself — prompt, brain, tools, agents, gates.
Every day at 08:10, run the daily standup
cronNew lead email → enrich + draft a reply
eventNew PR → review it across 6 rounds
webhookMRR drops 5% → investigate + report
signalcurl -fsSL sboghossian.github.io/side/install.sh | shThe real thing — not a mockup.
the canvas builds itself, then stops for your yes
every agent is a canvas — your whole fleet on one living map. Drag to pan.
Side is for anyone who uses AI every day but has never seen inside it, and for the people who practically live in there. No code needed, no ceiling either. You ask, you watch, you approve what matters. After a week you understand the thing everyone keeps talking about. That’s why I built it.
github.com/sboghossian/side·MIT·open source
Answer three things. Side assembles itself while you do.
Your agents read this before every run, so the work sounds like you.
Side runs on your own model key. It stays in this browser and is sent only to your model’s API.
Where should your brain start?
Full folder discovery deepens with the v0.8 engine — the brain keeps learning as you work.
Every agent is a canvas. Click one to open it. Some are public on your profile.
Head of Growth at a legal-AI startup. I ship websites, decks, outbound and reports with a fleet of agents. Plain language, fast results.
Live spend, budget caps, and where the tokens go — nothing phones home.
What runs while you sleep. Toggle it, see the next run, or run it now.
Every service and machine, live. The watchdog restarts what dies.
Install skills, agents, and fleet rails. Every plugin is SHA-256 verified.
Standups, cost, heartbeat, and weekly digests as a filterable newsroom.
Every action is append-only logged. Your keys, your machine, nothing leaves.
Every block is a full agent session. Trigger, input, model, agent, tool, logic, output.