About Provision

Managing one AI agent is fine.
Managing a team of them isn't.

We're building the operating layer for AI teammates — runtime, browser, email, and channels — so you stop being your team's sysadmin and start treating them like coworkers.

Origin

The Mac-on-the-shelf moment.

Provision started on a Mac in my home office. One agent. A Slack channel. A real inbox. It worked beautifully. So I built another. And another. Within a few weeks I had a team of agents running on a single laptop — a marketing lead, an SDR, a research analyst, a personal assistant — and a fan that never stopped.

Then I tried to scale it. Browser sessions started fighting for memory. Email deliverability got tangled. One agent's channel credentials would override another's. The Slack bot kept timing out because the laptop went to sleep. Every day I'd spend an hour as the team's reluctant sysadmin instead of working with them.

The realization was simple: managing one OpenClaw agent is fine. Managing a team of them is a job nobody wants. The runtime is solved. The browser is solved. The email is solved. What wasn't solved was running all of it together, reliably, for actual teams.

So I built Provision. We host the OpenClaw runtime, give every agent a sandboxed Chrome browser and a real email inbox, wire up Slack and Telegram and Discord with one click, and update the whole thing while you sleep. The open-source core is yours forever — we charge for the operations.

What we believe

A small set of strong opinions.

Open source is the only sensible foundation.

The agent runtime running your team should be inspectable, forkable, and self-hostable. We commit to OpenClaw and provision-core staying open-source under MIT — forever, not as a phase.

Agents are coworkers, not features.

A useful agent has a name, a role, an inbox, channels you talk to it in, and memory of what you've discussed. Anything less is a chatbot pretending to be an employee.

The channel matters more than the model.

Whether your team uses Claude or GPT or a local Llama is a config detail. Whether your agent shows up in the Slack you already check is the difference between useful and forgotten.

Hosting is a feature, not the product.

We don't compete with OpenClaw — we run it for you. The same way Vercel runs Next.js or Heroku ran Rails. Use Provision when you don't want to be the ops person; self-host when you do.

Default to honest pricing.

$99/month, flat. 48-hour free trial. We don't mark up tokens you bring. We don't charge per-seat for AI teammates the way SaaS charges for humans — that math doesn't apply when the teammate is software.

Work in the open.

Our roadmap, decisions, and post-mortems live on GitHub. The business that runs your AI agents shouldn't be a black box.

Founder

Who's behind this.

OA

Obaid Ahmed

Founder

Entrepreneur and technologist focused on the operating layer for AI teammates. Previously founded OAK Computing, TourZap, and UmmahHub. Carleton University grad. Top 40 Under 40 (Ottawa Business Journal & Chamber of Commerce, 2015). Splits time between the Bay Area and Ottawa.

Provision is a small, deliberate team — we're hiring as the load demands it, but the bar is high and we'd rather ship one strong person every six months than scale a roster. If running an OpenClaw cloud sounds like your idea of a great way to spend a few years, say hello.

Where

Built between two cities.

Provision was started in 2026, built between the San Francisco Bay Area and Ottawa. Most of the customer-facing work happens on US time; engineering tends to drift toward Ottawa hours. We're a remote-first company by default.

San Francisco Bay Area

Customers, partners, GTM

US

Ottawa, Canada

Engineering, infrastructure

CA

Open source

The core stays free, forever.

We build on OpenClaw. The Provision platform itself is open-source under MIT — github.com/provision-org/provision-core. You can self-host the entire thing for $0 in software fees; you'll pay for your own VPS and LLM API. Provision the product is the managed cloud version: same code, plus the infrastructure, deliverability, browser sandbox, and channel OAuth handled.

We don't do open-core gating — meaning we don't hold back features behind a license boundary to push you to the cloud. The features in Provision the cloud are operations features (managed updates, multi-tenant isolation, billing, SLAs, support) that don't make sense to open-source even if we wanted to.

Built with

The tools we credit.

Provision runs on a small, opinionated stack. The list below is the short version — the people who built these tools deserve the credit, and reading the stack tells you a lot about the engineering philosophy.

Laravel 12

Backend

React 19 + Inertia

Frontend

TypeScript 5.7

Types

Tailwind v4

Styling

Next.js 16

This site

OpenClaw

Agent runtime

Hetzner / DigitalOcean

Infrastructure

Anthropic + OpenAI + Ollama

Models

Stripe

Billing

Talk to us

Get in touch.

Questions, feedback, partnership ideas, hiring inquiries — one address handles all of it.

For privacy requests: privacy@provision.ai · For legal: legal@provision.ai

Hire your AI team.
We'll run the office.

$99/mo. 48-hour free trial. No credit card.