An operating system for running a business with AI. One shared engine of business abilities — customers, messages, documents, money, memory — that AI agents and full software products both plug into.
Most software makes you the operator: you click, it records. SLOE OS inverts that. Agents operate the business on your behalf, and every action they take — sending an invoice, paying an employee, filing a lead — runs through one secure, shared substrate.
Every ability is a small, named business action. Anything that can make an HTTP request — an app, an agent, a workflow — can invoke one. Access is governed by scoped keys per tenant: a key can only see and call what it was granted, and client data never crosses tenants.
One engine, two doors. Both hit the same abilities, the same code, and the same tenant isolation — your key decides who you are and what you can touch, whichever door you use.
POST https://substrate.sloelabs.com/abilities/execute
Authorization: Bearer <your key>
{ "ability": "message.send",
"input": { "to": "...", "channel": "email",
"message": "..." } }
Any app, backend, or workflow that can make an HTTP request. A client with an existing system integrates in an afternoon — no SDK required.
https://substrate.sloelabs.com/mcp
Paste one URL into Claude, Cursor, ChatGPT, or any MCP-capable agent and the abilities appear as native tools it can call directly. Tools outside your key's scope aren't just blocked — they're not even listed.
MCP is a thin protocol layer over the same substrate: a tool call through MCP and an HTTP call through the API run identical code with identical scoping. Products call the API; agents speak MCP; the engine doesn't care which.
A 24/7 agent on Telegram or WhatsApp handles the morning brief, drafts proposals and emails, chases follow-ups, files leads, and generates documents — you approve, it executes.
Sebenza runs SMBs end-to-end — quotes, invoices, payroll, HR, inventory, client portal — configured per industry (26 supported) and built for African markets: South Africa, Mozambique, DRC, Botswana.
AI scouting with computer vision, academy and club management, agency operations. Proven across ScoutBase, Sportnaa, and a live club deployment for Real Zaragoza's academy.
Customers pay invoices directly into the business's own bank account; staff get paid from it. SLOE never holds the money — it operates the rails and takes a small platform fee.
Proposal, invoice, contract, or scouting report: generated as a polished PDF from live data and delivered by email or chat — one ability call, no design work.
The substrate is a remote MCP server. Connect Claude, ChatGPT, Cursor — any modern AI client — and it can operate your business, seeing only the abilities its key allows.
A structured audit of your operation: where AI pays off first, what to automate, what to leave alone. Delivered as a concrete, costed roadmap in days.
A custom vertical OS assembled from proven abilities rather than written from scratch — which is why builds take weeks, not months.
SLOE operates the system with you: a dedicated agent, monitoring, monthly reporting, and continuous improvement as your business changes.
A new ability — say, a payout rail — lands in the substrate once and every product and agent inherits it immediately. No per-app rebuilds.
Self-hosted infrastructure under SLOE's control, with a roadmap to fine-tuned in-house models — so a client's data can ultimately live and compute on hardware they own.
Every key is scoped. A tenant can't call outside its scope — and can't even list what exists beyond it. An out-of-scope ability looks identical to one that doesn't exist.
Payments and payouts run through each client's own accounts. SLOE orchestrates the flow and charges a platform fee — funds never sit with us.