Work went project-based and agent-driven. The infrastructure that pays for it didn't follow. So we built it: chat-native project finance on regulated open banking and stablecoin rails, where one client invoice flows in and many compliant payouts flow out, callable by people and AI agents alike.
A single client payment fans out to every contributor, each settled on the rail that's fastest and cheapest for them. Watch it move.
Read and act on your project finances from anywhere. It all funnels into one Petl core that handles multi-party invoicing, then settles every contributor on the right rail.
Modern teams are assembled from contractors, subcontractors, fractional specialists, and now AI agents. They ship outcomes, not timesheets. But the money rails underneath still assume a salaried employee, or a single freelancer with one account. Splitting a client payment across a project team, staying compliant, and settling across borders is still slow and manual. As agents start doing real, billable work, that gap becomes a wall.
Petl is chat-native project finance. Work is logged from WhatsApp, Slack, Claude or ChatGPT. Petl turns it into one clean client invoice, splits it across every contributor, and settles each on the right rail. People and AI agents can run the entire flow, from logging work to triggering payouts, over MCP.
Underneath Petl sits regulated financial infrastructure that we built and own. We engineered the backend, the orchestration, and the compliance layer so a single invoice can fan out to many contributors, each paid on whichever rail is fastest and cheapest, automatically.
Instant, account-to-account settlement across the UK and EU, with no card networks in the middle.
// instant bank-to-bankNear-instant cross-border payouts to USD, EUR, ZAR and across LATAM, on regulated rails.
// near-instant cross-borderWorkOS authentication with permissions enforced at the rail, so every action is authorised before money moves.
// WorkOS · authExposed as an MCP server and deployed as a Claude connector and ChatGPT app, so people and agents reach Petl where they work.
// MCP · connectors · appsThis is what makes Petl built for what's coming. We embedded the payment mechanisms directly into the product and exposed them through an MCP server with WorkOS authentication. A person or an AI agent can, in natural language, list projects, check a wallet balance, and query a Pay by Bank or stablecoin payment link, without leaving the chat. The agent isn't simulating a payment. It's talking to real, regulated rails. Money movement becomes something software can request as easily as a database query.
Petl didn't start as a pitch deck. Rafiki Works delivered 55+ projects through market exploration and hands-on execution of payments across the contractor economy and distributed teams, validating the problem before a line of Petl was written. Petl has since supported over 30 international projects within its first three months of private beta. It's backed by the Visa Fintech Accelerator and the Baobab Network, with SEIS Advanced Assurance confirmed.
As more work is scoped as projects and more of it is executed by agents, something has to be the rails that pay for outcomes, compliantly, anywhere, callable by software. That's what Petl is building toward: canonical payment infrastructure for project-based and AI-driven work. The Stripe for the way work happens next.
Coming next: native tax calculations and a direct HMRC integration, so filing happens where the money already moves.
Petl is what happens when our team takes an idea all the way: real rails, real compliance, callable by AI.
Visit Petl Pay