The AI agent harness for payment teams
Trusted by teams running mission-critical operations.
Channels
Apps
Maya 9:12 AM
re-review Harbor Vape Supply, dispute ratio spiked overnight
APP 9:13 AM
Dispute ratio is 1.4%, above the 0.9% VAMP threshold. KYB still matches, but the website changed owners on Sep 22 and 61% of disputes are "not received".
Recommend a 25% rolling reserve. 14 queries, 3 sources.
Approval needed
Apply 25% reserve to Harbor Vape Supply
DK2 repliesLast reply today
Dev 9:15 AM
Can you check their two sibling MIDs too?
Message #risk-ops
Digital teammates, built by anyone on your team
Describe the job in plain English. Your agent gets its own computer, inbox, and phone number, follows your procedures, and asks before it acts.
How Runtime works
Describe the agent
Anyone on the team can describe a job in plain English. Add your procedures and the points where a person has to approve.
Create a banking ops agent that handles RFIs, hold-harmless requests, and returns of funds from our partner bank, verifies against our ledger, and drafts replies for my approval.
bank-ops-agent
Tools picked from your prompt
StripeConnect your existing tools
Give it the systems it needs, with role-based access that sets what each agent and teammate can read and do.
Approved systems and permissions for this agent.
Sensitive actions wait for your team’s approval.
Put it to work
Call it from Slack, Teams, email, or phone, or trigger it from an alert or schedule. It works on its own computer and carries out the steps you allow.
Column flagged a transfer on account 4417 and is requesting a return of funds. Can you take it?
Looking into it. Reviewing the request:
Every step visible while it runs. Cost so far: $0.19
Review and approve
Approve sensitive actions and inspect the evidence, activity, and outcome of each run.
Return $12,400 to the originating bank for the flagged transfer, with a drafted reply to Column and the evidence file attached.
Built for payment ops and risk teams
Payment ops, underwriting, fraud, compliance, and support teams use Runtime to investigate cases end to end. Agents pull the evidence from every system, draft the outcome for sign-off, and resolve the escalations that would otherwise land on engineering.
Payment operations
Clear reconciliation breaks, exceptions, and proof-of-authorization requests before they turn into a backlog and a headcount request.
Learn morePayments support
Investigate declines and payout issues using real transaction data, and deflect escalations from engineering.
Learn morePartner-bank requests
RFIs, hold-harmless requests, and returns of funds from your partner bank. Agents parse the file, investigate, and draft the reply.
Learn moreUnderwriting, KYB and KYC
Underwrite new merchants and re-review your portfolio on a schedule. Agents assemble the evidence so your team makes the call.
Learn moreFraud and transaction monitoring
Triage alerts, cut false positives, and trace transaction patterns, with an evidence trail for every query the agent ran.
Learn moreCompliance
Alert investigations and card-network programs with the evidence trail regulators and partner banks expect.
Learn moreRun sensitive payment data inside your own cloud
Every agent works on an isolated computer you control, with human approval on the actions that matter.
How we handle securityComputers in your cloud
Provision agent computers in your own AWS, GCP, or Azure account, or self-host the whole platform. PCI data stays inside your environment.
Keep card data out of the model
Agents use masked credentials and never see raw secrets. Hooks and filters you configure keep card numbers and PII out of prompts and logs.
Role-based access
SSO for the whole team. Each agent and teammate gets only the tools and data their role allows, and money movement waits for approval.
SOC 2 compliant
Independently audited against the AICPA Trust Services Criteria. Encrypted at rest and in transit. We never train on your data.
A system of intelligence for payment teams
Runtime is the system of record for the work your agents do and the system of action that does it. Every run is stored, and every run makes the next one cheaper and faster.
The request, every query and tool call, the approvals, the cost, and the result. Your agentic system of record, ready when compliance, an auditor, or your partner bank asks.
Re-review Harbor Vape Supply against VAMP thresholds
Tool calls
14
Duration
4m 02s
Cost
$0.31
Stored, searchable, and exportable.
Agents learn from resolved cases and propose updates to their own skills, so your runbooks improve as your team works.
Learned from the last 40 resolved cases:
Suggested skill update
Add step 4: check sibling MIDs before closing a re-review.
Runtime is a neutral provider that works with any model and any harness. Evals find the smallest model that still passes each skill, so you stop paying frontier prices for routine work.
Cost per run vs. the original model. Illustrative.
Better skills mean fewer steps, smaller models mean fewer dollars, and sandboxes pause when idle and boot from snapshots. See token and compute cost by agent, team, and run.
Every run stored, searchable, and auditable
A partner for your AI agent rollout
Our forward-deployed engineers work with your team to map processes, build the first agents together, and set up guardrails, so your team can build the rest. We’ve built payment infrastructure and AI systems at companies like these.
How we work with payment teams
Frequently asked questions
Who is Runtime for?
Does card data or PII ever reach the model provider?
Can we run Runtime in our own AWS, GCP, or Azure account?
We already built internal bots. Why not build this ourselves?
Will it work with our existing and legacy systems?
How do approvals and the audit trail work?
Which models can we use, and how do you keep costs down?
What does a pilot look like?
See Runtime in action
See how agents work your payment operations and risk queues inside your own cloud, with your team approving what matters.