What Firmly actually sells
Firmly closed a 15 million dollar Series A led by Squarepoint Capital, with Mastercard, J.P. Morgan Payments, and Samsung NEXT also participating. The company's pitch is narrower and more useful than most agentic commerce messaging: it is an integration layer, not a storefront. Its flagship product, firmly Connect, is built so that a merchant plugs in once and becomes sellable across AI shopping agents, marketplaces, social platforms, and shoppable ad formats, rather than negotiating and building a separate connection to each one as it emerges.
The company also shipped Ask Agent, a conversational shopping widget merchants run on their own websites, pairing natural language product search against a live catalog with in-chat checkout and support. Co-founder and CEO Kumar Senthil put the merchant problem plainly: retailers cannot build a new integration every time a new storefront shows up. That is the argument for buying this layer rather than building it, and it is the argument every retail CTO fielding agentic commerce requests this year needs to pressure-test before committing engineering time.
Why payments players, not just VCs, wrote this check
Mastercard and J.P. Morgan Payments are not typical Series A investors in a commerce tooling startup, and their presence here is a tell. Payment networks make money on transaction volume and want agentic commerce to route through rails they can see, price, and secure, rather than through opaque browser automation they cannot audit. Backing an integration layer that sits between merchants and agents gives them visibility into how agent-initiated purchases actually clear, which matters enormously once chargeback, fraud, and liability questions around AI-initiated transactions start showing up in earnest.
Co-founder and President Damien Balsan described firmly as an infrastructure layer combining browser automation with deterministic protocols such as the Universal Commerce Protocol, built so merchants do not need to write new code to participate. That framing, infrastructure rather than application, is exactly what a payments network wants to underwrite, because infrastructure consolidates and compounds in value while individual shopping experiences get replaced every product cycle. For a CFO evaluating the same vendor, the practical read is that Mastercard and J.P. Morgan's involvement buys the startup credibility with risk and compliance teams that a pure venture round would not, which shortens procurement cycles for retailers already anxious about agent-initiated fraud.
The build versus buy question this creates for merchants
Every retail CTO now fields some version of the same request from the business side: get us into ChatGPT's checkout, get us listed in Gemini's shopping agent, get us onto whatever marketplace launches next quarter. The honest build option means maintaining a growing set of bespoke integrations against protocols that are still being finalized, each with its own authentication, catalog feed format, and settlement logic. That is a real and growing maintenance burden, and it only gets worse as the number of relevant agent surfaces increases.
Firmly's bet is that most retailers will conclude, correctly, that this integration sprawl is not a strategic asset worth owning internally, the same way most retailers stopped building their own payment gateways once PCI compliance and card network certification made that a specialist's job. The open question is whether firmly Connect becomes the de facto layer before a hyperscaler or a payments incumbent builds a competing version natively into its own checkout stack, which would make a standalone vendor's independence either a selling point or a liability depending on how the market consolidates.
Governance and risk CTOs should price in now
Routing purchase intent through a third-party layer that touches every emerging agent surface concentrates risk in a way procurement and security teams should scrutinize closely. If firmly Connect sits between a retailer's catalog, pricing, and checkout and every AI agent that can initiate a purchase, its uptime, data handling, and access controls effectively become the retailer's own. Contract terms should specify data residency, audit rights over transaction logs, and clear liability allocation when an agent-initiated order is disputed, since the deterministic protocol layer Balsan describes still needs a human-auditable record for compliance and fraud teams.
There is also a product data dependency worth naming. Any layer that lets AI agents discover and purchase products is only as good as the catalog feed behind it, and retailers with messy product data, inconsistent pricing rules, or thin inventory feeds will get poor agent visibility no matter how good the integration layer is. Firmly's pitch solves the connection problem, not the underlying data quality problem, and CTOs should budget for catalog cleanup as a prerequisite, not an afterthought.
Where this fits in the broader agentic commerce land grab
This round lands alongside a wave of similar funding activity across agentic commerce infrastructure, from catalog optimization tools to attribution platforms trying to solve the problem of knowing which AI agent actually drove a sale. Firmly's angle, a single connection point rather than a single feature, positions it to be useful regardless of which specific agent or marketplace wins the most consumer attention, which is a defensible place to sit if the agent landscape stays fragmented through 2027.
The risk is the opposite scenario: that OpenAI, Google, and Amazon each consolidate enough commerce volume that merchants only need to integrate with two or three dominant surfaces directly, making a neutral connector layer less necessary than it looks today. Firmly's near-term growth, and whether named retail customers emerge beyond the launch announcement, will be the clearest signal of which scenario is actually playing out, and procurement teams should ask any competing vendor the same question before signing a multi-year infrastructure contract on a market this young.
What CTOs should do before the next pitch meeting
Treat any agentic commerce integration vendor, Firmly included, as infrastructure procurement rather than a marketing tool purchase, which means routing the evaluation through security, payments, and data governance teams before it reaches a marketing budget line. Ask for specifics on which protocols are actually supported today versus roadmapped, since the agentic commerce standards landscape, including UCP and competing approaches from OpenAI and Google, is still consolidating and vendor claims outrun what ships.
Finally, insist on contractual portability. The value of a connection layer evaporates if switching vendors later means rebuilding every integration from scratch, so exit terms, data export guarantees, and catalog feed ownership deserve the same scrutiny as the sales pitch about zero engineering effort. A 15 million dollar Series A buys Firmly runway to prove the model, not yet the scale that makes vendor lock-in a safe bet for a retailer's checkout stack.
