Company Spotlight: Beltic Builds Trust for AI Agents

Beltic is building trust infrastructure for AI agents. The San Francisco company gives websites, APIs, and MCP servers a way to identify an agent and its operator, understand what the agent is trying to do, and apply policy before that action reaches a production system.

Co-founders Isha Bhatnagar, CEO, and Mike Allan Pellin, COO, launched Beltic in 2025. CTO Farhan Afsahi and CRO Luca Castellano round out the current leadership team. Their target market sits where identity, security, compliance, and agentic commerce collide, which is usually where software diagrams become risk committee meetings.

Beltic matters now because AI agents are moving from conversation to action. They can search, book, pay, call tools, and invoke other agents. Every new capability creates a corresponding question for the business receiving the request: who sent this software, what authority did it receive, and should this particular action be allowed?

About Beltic

Beltic describes its product as a control layer for incoming agent traffic. The platform combines 3 functions: agent visibility, a policy engine, and intent intelligence. Visibility identifies the agent, its operator, and its origin. Policy defines which accounts, actions, and resources carry risk. Intent intelligence evaluates the real-world action behind a technical request.

That last piece matters. A server can authenticate a credential and still misunderstand the authority behind a transaction. A valid login does not prove that software was allowed to move $1,000, change a booking, retrieve sensitive data, or hand the task to another agent. Beltic is trying to evaluate the identity and the action together.

The company's broader product stack includes Know Your Agent credentials, business verification, user verification, document intelligence, workflows, and compliance case management. Know Your Agent uses cryptographic credentials tied to a verified person or business, giving a receiving platform a way to verify trust without exposing all of the underlying identity data.

Why Beltic Matters Right Now

Agentic commerce is becoming an infrastructure market. Visa's Trusted Agent Protocol helps merchants recognize approved agents and validate signed messages. The Agentic Commerce Protocol creates a programmatic interface among buyers, agents, and sellers. Those systems help traffic move. Beltic is focused on the decision surrounding the traffic.

The distinction is easy to miss when an agent books a restaurant table. It becomes impossible to ignore when an agent initiates a payment, accesses a regulated workflow, or calls another agent carrying borrowed authority. Machine-speed transactions compress the time available for review, while chains of delegation make accountability harder to trace.

Beltic's thesis is that identity, authorization, intent, and policy must remain connected across those handoffs. That is a more ambitious job than bot detection. It asks the receiving business to judge not only whether automation is present, but whether that automation is legitimate for the action in front of it.

Leadership Built Around Regulated Systems

Beltic's founding team has spent years inside the systems it now wants to connect. Bhatnagar spent 6 years at Coinbase working on global regulatory experience, identity, onboarding, and early institutional customers. Allan Pellin previously founded Atar, a Brazilian banking-as-a-service company acquired by Porto Seguro in 2021.

Afsahi brings the data layer. He co-founded Verifiet, the data-infrastructure company Beltic acquired in 2025. Beltic and lead investor Norwest say that acquisition brought direct jurisdictional connections and coverage of more than 500M entities. The number is company and investor reported, not an independently audited operating metric, but it explains the architecture: an agent credential is only as trustworthy as the principal verification underneath it.

Beltic has disclosed $8.8M in funding. Its latest $7.3M Seed was led by Norwest, with Restive Ventures, Oxford Seed Fund, and Collide Capital participating. The company has not disclosed revenue, valuation, or named production customers.

Product Discipline and Security

Beltic's published operating principles are unusually specific for a young infrastructure company. The team says it chooses boring, predictable technology where reliability matters, keeps customers directly involved in product decisions, maintains a small product surface with deep support, and surfaces bad news while it is still cheap to fix.

Those principles fit the category. Trust infrastructure does not earn points for surprise. It must produce decisions that engineering teams can implement, security teams can inspect, and regulators can reconstruct. Beltic says a SOC 2 Type II report is available, conducts annual penetration testing, encrypts data at rest and in transit, and is pursuing ISO 27001 certification. These are company disclosures, but they set a concrete standard against which customers can evaluate the platform.

Beltic also publishes case studies across insurance, stablecoin onboarding, global user verification, and legal claims. The customer identities and quantified outcomes are not independently verified, so they should be read as product evidence rather than proof of broad commercial adoption.

What Beltic's Hiring Signal Says

Beltic currently lists 2 open roles: Go-to-Market Lead and ML Engineer. The combination is more revealing than a generic hiring announcement. One role is responsible for reaching platform and infrastructure teams already seeing agent traffic they cannot explain. The other is responsible for models that distinguish trusted agents from attacks under real-time load.

That hiring pattern suggests Beltic is trying to prove 2 things at once: that the technical decisioning works and that the market is ready to buy an independent control layer. The company describes itself as a small, senior, remote-first team across UTC-5 to UTC+3, where everyone talks to customers and everyone ships.

The opportunity is real, but so is the adoption challenge. Platforms may build controls internally, payment networks may extend their protocols, and identity or cybersecurity incumbents may absorb agent verification into existing products. Beltic must show that its integrated data and policy stack produces better decisions without becoming another system customers have to reconcile.

The Bigger Shift in Agent Infrastructure

The agent economy is developing the same way most software markets do: capability arrives first, controls arrive after somebody discovers that a demo has no audit trail. Beltic is entering during that handoff from possibility to accountability.

Its strategic edge is not the claim that agents need identity. That is quickly becoming consensus. The stronger claim is that identity alone is insufficient. A useful trust layer must understand the principal, the agent, the delegated authority, the requested action, and the policy of the system receiving it.

Beltic will be judged by production evidence: integrations, repeatable customers, decision quality, latency, and the ability to work across protocols. If the company succeeds, agent verification will stop looking like a niche compliance feature and start looking like a required control plane. If it fails, businesses will keep choosing between blocking automated traffic and letting unfamiliar software knock on the door wearing someone else's badge.