Runtime and Middesk

Middesk and Runtime both come up when a B2B fintech or payments company wants onboarding and merchant reviews to stop piling up. They do different jobs. Middesk tells you who a business is, from authoritative sources, and now has agents that resolve identity questions. Runtime agents take the case the rest of the way, across your own systems and the teams that have to act. Most teams would run both.

What each does

Middesk is business identity infrastructure. Its Verify product checks a business's name and TIN against the IRS, searches Secretary of State filings across all 50 states and DC, screens the business and its people against OFAC and other watchlists, and monitors businesses for changes. It also offers underwriting data like UCC filings, embedded business registration, and fraud models. It now also verifies businesses in 30 countries across Canada, Europe, and APAC through the same API. Middesk Agents are API-first agents for CIP, CDD, EDD, merchant risk, and fraud investigations: they resolve OFAC false positives, address discrepancies, beneficial ownership, litigation research, and operating status, with full evidence on every finding. Middesk says it verifies 7 million businesses a year.

Runtime is the AI agent harness for payment and fintech teams. Anyone on risk, compliance, onboarding, underwriting, payment ops, finance, or support builds agents from their SOPs. Agents work on their own isolated computers in your cloud and reach tools through APIs, databases, CLIs, MCP servers, and a browser for portals with no API. They start read-only, stop for approval before anything moves money, and every run is recorded end to end.

Put simply: Middesk answers "is this business real and who owns it." Runtime agents answer "what do we do about this merchant," with every team's data.

Where they overlap

Middesk Agents are the honest overlap. Middesk pitches them as a way to skip building manual review queues for KYB, with audit-ready decisions that don't reach your compliance team. For a pure identity question, like whether an OFAC hit is a false positive or whether a registered address matches, Middesk Agents work with Middesk's own data and are a natural first choice.

The overlap ends where business identity ends. A merchant review at a payments company is rarely only KYB. It is also processing volume against what the merchant said at onboarding, a spike in chargebacks, a reserve decision, a held payout, an angry email to support, and a note to the sponsor bank. Middesk doesn't hold that data and isn't built to act on it. That is the work Runtime agents do.

How they work together

A typical flow for merchant onboarding exceptions and periodic re-reviews:

1. Middesk verifies. A new business is submitted to Middesk, or Middesk monitoring reports a change: a dissolved registration, a new owner, a watchlist hit.
2. Middesk Agents resolve what they can. Identity questions, like a false-positive OFAC match, come back with a decision and evidence.
3. A Runtime agent takes the rest. Through Middesk's API, webhooks, or a Slack alert, the agent pulls Middesk's findings and then checks your systems: processing history in your ledger, disputes and payouts at the processor, underwriting notes, and support tickets. It uses the browser for portals with no API.
4. It drafts the decision with evidence. Following your SOP, the agent writes the case: what triggered it, what Middesk found, what your own data shows, and a recommendation such as approve, request documents, adjust limits, apply a reserve, or offboard.
5. A person approves in Slack or Teams. Applying a reserve, holding a payout, or replying to the sponsor bank always waits for a person.
6. The agent acts and closes the loop. It updates the merchant record, notes the ticket, tells support what to say, and keeps one record of the run: trigger, every query and tool call, approvals, cost, and result.

When a pattern repeats, like a common registration mismatch, the agent can turn it into a script that runs deterministically as a template in your repos, owned by your engineers.

When you need both

• You onboard businesses or merchants and reviews need more than identity data
• KYB changes trigger work for underwriting, payment ops, finance, or support
• You run periodic merchant re-reviews against real processing behavior
• You want agents in your cloud, with models you choose, and approvals on money movement
• Your sponsor bank wants the full story of each merchant decision in one record

When Middesk alone is enough

• Your backlog is identity questions that Middesk Agents can resolve
• You don't have downstream merchant risk or payment ops work tied to KYB
• You only need verification and monitoring through one API
• Other teams aren't ready for agents

Frequently asked questions

Is Runtime an alternative to Middesk?

No. Middesk verifies businesses using data from sources like the IRS and Secretary of State offices, and monitors them over time. Runtime doesn't provide that data. Runtime agents use Middesk as a tool and do the review and follow-up work around it.

Does Middesk have AI agents?

Yes. Middesk Agents are API-first agents for business identity work, such as OFAC false positive resolution, address discrepancies, beneficial ownership, and CIP, CDD, and EDD investigations. Each finding comes with evidence and reasoning. The CIP workflow launched in public beta for existing customers.

What does Runtime add if Middesk Agents resolve KYB cases?

Middesk Agents work on business identity. Merchant and business reviews often also need your own data: processing volume, chargebacks, held payouts, support history, and underwriting notes. Runtime agents work across those systems and the teams that own them, and keep one record of the case.

How much does Middesk cost?

Middesk does not publish pricing. It is sold through sales.

Does Middesk cover businesses outside the US?

Middesk is US-first, with US Secretary of State coverage across all 50 states and DC. It now also verifies businesses in 30 countries across Canada, Europe, and APAC through the same API.