Turn your files, databases, and APIs into something AI can talk to
Ask most small businesses where their real operational knowledge lives, and the answer is rarely glamorous. It's a spreadsheet someone built three years ago and everyone's afraid to touch. It's a MySQL database quietly running an old inventory system. It's a third-party API some developer wired up once and nobody documented. None of it was ever meant to talk to an AI assistant. Now it can.
Hutif.com has expanded what it turns into a remote MCP server: not just uploaded files, but live databases and existing APIs too. The idea stays the same across all three - connect a source, get a hosted link, plug that link into Claude or another AI agent, and start asking real questions about real data. What changes is how much of a business's actual infrastructure now qualifies as "a source."
Three doors into the same room
The three input types cover most of what a small operation actually has lying around:
- Files. Upload a CSV or Excel sheet and Hutif reads the structure, generates the right tools, and exposes it - search, filter, add, update - without anyone writing a line of code.
- Databases. Paste a connection string for an existing Postgres, MySQL, or SQL Server database, and Hutif connects directly to the tables that are already there. No migration, no export - the AI reads and writes against the live system.
- APIs. Point Hutif at an existing API and it wraps the calls into MCP tools an AI agent can invoke on request, turning a piece of internal tooling into something conversational.
None of these require the person setting it up to know what an API is, let alone build one. The technical translation happens once, inside Hutif, so it doesn't have to happen again for every business that shows up with a slightly different mess of data.
Why the database and API pieces matter more than the file upload did
File upload was the easy win - a CSV is self-contained, static, low-risk. Databases and APIs are a different order of usefulness, because they're live. A connected database reflects what's true right now, not what was true when someone last exported a spreadsheet. An AI agent asking a live database "what's low on stock" gets today's answer, not last Tuesday's.
That immediacy is also where the responsibility gets heavier. A connection string to a production database is access to a production database. Hutif's job isn't just translation anymore - it's drawing the lines around what an AI agent is allowed to touch, and making sure a vague instruction can't quietly become a destructive one.
Who this is actually for
Not enterprises with dedicated integration teams - they already have ways to do this. It's for the accountant with a client database in MySQL, the small e-commerce operation whose inventory logic lives in a half-documented internal API, the agency with client data scattered across a spreadsheet and a database and a booking tool. The kind of setup that's too specific for an off-the-shelf integration and too small to justify hiring a developer.
The bet behind Hutif is that this segment is larger than it looks, and that almost nobody has been building for it - because "connect my weird internal database to AI" doesn't sound like a product category until you say it out loud.