Minicor Raises $3.8M Seed for Legacy Desktop API Access

A prospective customer with an on-prem electronic medical record system helped change the direction of Minicor. The founders had been working on observability and recovery for API integrations; the buyer's need drew them toward desktop automation. Their product now gives AI developers a callable route into applications whose daily work still happens through a screen.

The San Francisco company announced a $3.8M Seed round on October 14, 2026, led by Brad Gillespie at IA Ventures, with Y Combinator and Paul Graham participating. Co-founder and CEO Faizaan Chishtie and co-founder and CTO Saheed Akinbile are building infrastructure for AI vendors that need to connect to software their customers already operate.

That relationship puts an established application inside the economics of a new AI sale. Minicor offers a way to reach it without first asking the customer to replace it, while taking on the continuing work of running the desktop workflow. For developers, the useful promise is access to the existing business process, together with enough visibility to understand what happened when the process needs attention.

What Minicor's $3.8M seed supports

The company's announcement names Pioneer Fund, Robinhood Ventures, Garage Capital, Rebel Fund, Blast, SuccessVP, K5 Global, 20VC, Phosphor Capital and Metsu Capital among the other participants. It also identifies Austen Allred, Henri Stern, Matteo Franceschetti and Jeff Henriod. Those names establish the disclosed investor group; the announcement leaves individual allocations and valuation undisclosed.

Minicor is hiring, with its careers page describing infrastructure work across real Windows environments. The financing accompanies a product already addressing legacy desktop access, although the company has not provided a detailed spending split. Treating the round as a precisely budgeted expansion plan would add specificity the announcement does not supply.

The founders' account on Y Combinator gives the commercial history more texture. They describe a prospective EMR connection that pulled the company toward desktop automation after earlier work on API integrations. A buyer's existing installation became an instruction about where their product needed to go, bringing the founders closer to the software an AI vendor would have to support in production.

That is an important choice for a young infrastructure company. A desktop interface creates a continuing dependency on behavior outside its own codebase. Taking that dependency into the service gives a customer another supplier to evaluate, while giving the AI developer a possible alternative to maintaining each connection internally.

How an API call reaches a desktop application

In Minicor's documentation, the unit of automation is a job with an endpoint. Developers teach the job through sample API inputs, and the builder refines the automation against those examples. A Windows desktop service performs the application steps, allowing a product to request a workflow through a familiar programmatic interface.

This arrangement keeps the older application in the working process. Its screens, fields and sessions remain part of execution even when the developer calling the endpoint does not interact with them directly. That separation is commercially useful because the buyer can pursue an AI application while retaining the system that already holds its operating records.

The distinction also explains why observability belongs close to the product's core. Minicor describes recorded runs, step logs, human review and status events among its tools. These are vendor-described capabilities, and their practical value depends on how a customer's own workflows behave; the documentation is not an independently measured reliability benchmark.

The company's use-case catalog spans tasks such as entering clinical information, updating business records and handling automotive workflows. Those examples show the kinds of connections Minicor is offering. They should be read as a product catalog rather than a list of publicly identified customers or proof that every application can be automated without additional work.

The integration work inside an AI customer relationship

An AI vendor selling into healthcare, financial services, automotive or logistics can encounter a customer whose useful data and required writes live in a desktop system. The buyer may like the proposed AI experience while still needing a dependable connection to that application. Integration then becomes part of the commercial commitment, affecting what the vendor can demonstrate, deliver and support.

Minicor is positioning its interface at that point in the relationship. For an AI developer, calling an endpoint can simplify the shape of the integration it maintains. The responsibility for the customer's outcome still crosses several organizations, making failure reports and recovery paths part of how those organizations work together.

DevCuration's coverage of RapidCanvas's managed enterprise AI examines the continuing engineering relationship around deployment. Hypercubic's mainframe modernization follows work around another established computing environment. Minicor addresses desktop execution specifically, placing its service beside the application a customer has already decided to keep.

A developer assessing that service needs to bring actual cases into the evaluation. Sample coverage, write approvals, failure handling and the evidence available after an interrupted run all affect the usefulness of the connection. The relevant comparison is the work required to operate the intended workflow through each available approach, including the support burden carried by the vendor and its customer.

Where the desktop connection keeps developing

The initial financing gives Minicor backing for a service with an unusually persistent neighbor: somebody else's installed software. A successful connection becomes useful precisely because that software remains in the customer's operation. Changes to its screens or sessions can therefore become part of Minicor's own working day.

The company has described mechanisms for recording, controlling and recovering those workflows, and it is recruiting people to keep building them. Public materials do not establish an audited customer count, universal success rate or detailed allocation of the seed proceeds. Buyers can use the disclosed mechanics to frame an evaluation without turning those missing figures into assumptions.

For Chishtie and Akinbile, the EMR prospect that redirected their work illustrates how this business can keep taking shape. An AI vendor arrives with something new to sell, and the customer points toward an application already doing necessary work. Minicor is building the connection those two parties need to keep using together.