The platform is being developed to give AI agents greater control over payment operations while maintaining defined mandates, compliance checks and settlement processes.
Mesta has launched Agentic Payments, an infrastructure layer that enables businesses and AI agents to interact with its global payment network.
Phase One, available now, introduces a Model Context Protocol (MCP) server designed to simplify enterprise integration with Mesta. The company has also set out a three-phase roadmap that will extend the platform from connectivity to governed payment workflows and, ultimately, end-to-end agentic transactions within limits a named person has approved in advance.
Agentic Payments is built on Mesta’s global fiat and stablecoin settlement network, which has processed more than $2.5 billion across over 40,000 transactions in 22 months.
Mesta’s approach reflects a core principle: an Agentic payment system requires established payment infrastructure underneath it. AI agents can interpret instructions and coordinate workflows, but they still need settlement rails, authorisation controls and compliance processes to move money.
“Agentic payments cannot operate as a software layer disconnected from the underlying payment network. We built the settlement rails first and proved them at scale. Today, we are opening those capabilities to AI systems through MCP and setting out a clear path toward governed Agentic payments,” said Sandeep Pyapali, Founder and CEO at Mesta.
Phase One: Connecting AI agents to Mesta through MCP
The first phase of Agentic Payments gives enterprises a simpler way to connect their AI systems with Mesta’s payment capabilities.
Instead of integrating separately with a series of payment API endpoints, businesses can use Mesta’s MCP server as a common interface through which agents can discover available payment functions, read quotes and payment status and submit payment instructions that are held until a named person releases them.
Based on implementation estimates, Mesta expects this approach to reduce some enterprise integration timelines from several weeks to a matter of hours. Actual timelines will depend on each organisation’s systems, requirements and approval processes.
“Phase One gives AI agents a structured way to interact with our payment infrastructure. The next steps are about allowing those agents to configure workflows and execute transactions while remaining inside the permissions set by the business,” said Kiran Polavarapu, Head of Engineering at Mesta.
A roadmap from connectivity to governed Agentic payments
Alongside the Phase One launch, Mesta is outlining the next two phases of its Agentic Payments roadmap.
Phase Two: Business workflows and mandate controls
Beginning in Q4 2026, Phase Two will introduce business process workflows accessible through Mesta’s APIs and MCP server.
These workflows will allow client agents to configure processes such as payroll, invoice settlement and liquidity movement without requiring the business to build each step through a direct API integration.
Phase Two will also introduce Mesta’s mandate layer. Businesses will be able to define what an agent is permitted to do, including spending limits, expiry periods, approval requirements and off switches. Limits are set at the organisation level and apply to every agent acting on its behalf, so no combination of agent permissions can exceed them.
Phase Three: End-to-end Agentic transactions within approved limits
Phase Three will support end-to-end agentic transactions within limits a named person has approved in advance, with anything outside those limits waiting for a person.
Mesta plans to expand the mandate layer to incorporate Know Your Business and Know Your Agent checks within a unified compliance verification step as agent identity standards mature. Know Your Agent is an emerging practice, not yet a ratified standard, for establishing and verifying the identity, authority and operating boundaries of an AI agent.
This phase is intended to ensure that authorised businesses and the agents acting on their behalf remain within assigned permissions and risk parameters throughout a transaction.
Four stages for every Agentic payment
Mesta’s Agentic Payments architecture is built around a four-stage framework:
Intent: Mesta records the outcome the agent has been instructed to deliver, such as running payroll or settling an invoice.
Mandate: Mesta checks the request against the organisation’s permissions, spending limits and operating rules.
Authorisation: A named person approves the workflow and its limits once and authorises anything outside them. The authorisation covers the exact payment plan.
Settlement: Once the required checks and approvals are complete, funds move through Mesta’s payment network and the transaction record is written.
Under this framework, when a business’s agent is instructed to run payroll, it drafts the payroll request and submits it to Mesta. Mesta compiles the payment plan and checks it against the organisation’s mandate and sanctions lists. A named person approves anything that requires human authorisation. The payments then settle and a tamper-evident record is created. Every check runs on Mesta’s side, outside the agent and its model, so an instruction to the agent cannot lift a limit.
The framework is designed to help organisations integrate through MCP, automate internal workflows and authorise internal or third-party agents to execute transactions within defined controls.
“Businesses should not have to choose between automation and accountability. People set the permissions and risk boundaries. Agents work within them. Every payment retains a record of what was requested, what was allowed, who approved it and how it settled,” said Sandeep Pyapali, Founder and CEO at Mesta.
Applying agents to payment operations and compliance
Mesta has also developed agents for payment operations and compliance.
The Payment Operations Agent monitors system health, identifies payment issues and gathers information across systems in read-only mode. It can recommend responses based on cost and risk, including repeatable actions such as retries and status refreshes, for an operator to carry out. It does not move money.
In Mesta’s internal use, the agent has reduced the time required to diagnose a failed payment from more than 40 minutes to under 10 minutes.
The Compliance Agent rechecks entities when sanctions lists change, assembles evidence and flags missing information. Its recommendations are linked to their sources and recorded for audit. The agent does not approve customers and a named compliance officer remains responsible for sign-off.
Mesta says the Compliance Agent has reduced entity review time from about 90 minutes to under 15 minutes.
Mesta plans to publish four mechanisms supporting its Agentic architecture in October 2026. These cover signatures tied to exact payment plans, the sealing of large payment batches as a single object, independently verifiable records and the inclusion of sanctions data in the authorisation process.
The company is also preparing two provisional patent filings covering elements of the architecture.
