Think inside your AI world.
How do I connect Ramp to Claude?
Ramp's MCP server puts the spend surface — transactions, vendors, purchasing patterns — where you can question it conversationally, which suits a tool whose whole pitch is spend visibility.
Direct answer: connect Ramp's MCP server in Claude via Settings → Connectors (or 'claude mcp add' in Claude Code) and authorise. Claude then reads your spend: transactions, vendor adoption and purchasing patterns — the dataset Ramp builds across tens of thousands of businesses, scoped to yours. Ask where the money went this month and the answer is your card data, not conjecture.
The visibility you get immediately
Vendor-level spend, category drift, the subscriptions that renewed quietly, the outlier transaction worth a look — surfaced by asking rather than by filtering dashboards. Ramp's data model is tidy, which makes the conversational read unusually crisp.
The ceiling that isn't in the card data
Your call, say, is a per-seat software cap: tooling spend stays under £120 per person per month, ratified when three overlapping AI subscriptions taught the lesson. Ramp itemises every renewal. Whether the stack now breaches your cap — and that the cap exists because of the overlap incident — is context the card platform was never told.
Spend, read against the cap
Through Unl the month's tooling spend arrives divided by seats and held against your ceiling: £134 — over, with the overlap story attached as the reason the line exists. The spend review becomes a pass-or-breach read against rules you own, rather than a scan you perform. Ramp supplies the ledger; you supplied the law; Unl keeps them in the same room.
Ramp in Claude answers where the money went; Ramp through Unl answers whether where-it-went clears the ceilings you set — per rule, with the incident that created each rule still legible.
Reads through Unl arrive with measured context — in the presence of the decisions you’ve already settled. The reach lane is live: one box, paste anything. If it speaks MCP, Unl can reach it. Readings arrive unprompted, the data beside the criterion; Unl is a courier, not a warehouse, and keeps only your keys and the frame.
Read further
Questions people ask
What does Ramp's MCP server expose?
Spend and transactions, vendor adoption and purchasing patterns — your organisation's slice of the dataset Ramp maintains across 50k+ businesses. It is an observational surface suited to questioning spend, not moving it.
Can I use this from ChatGPT as well?
Yes — the server speaks MCP, so any MCP-capable client connects the same way. Unl likewise: it is fullest on Claude and Claude Code, where it is tuned, and reaches any MCP surface.
What kind of rules make sense to ratify over spend data?
Ceilings and caps with reasons: per-seat software limits, category budgets, a single-purchase threshold that triggers a second look. The pattern is always a number plus its why — that pairing is what turns a spend read into a verdict.
Can I just use Ramp's MCP server on its own?
Yes — for the raw read it stands up on its own, and this page's steps need nothing else. It returns your transactions, vendors and purchasing patterns accurately, as an observational surface suited to questioning spend rather than moving it. What it cannot tell you is whether a month's spend is acceptable, because acceptable is a ceiling you set and Ramp was never told the number or the reason. That is the layer Unl adds on top of the same read, not a replacement for it.
Does Unl change or store my Ramp data?
Not stored: what Unl reads is quarantined for the turn and never kept — no query text and no read content reaches our telemetry. Not changed by a read, either — a read only reads. Where a server exposes a tool that writes, Unl can call it, because the socket exposes a server's full surface rather than a chosen subset; but every write carries your explicit in-turn gesture, verified at the choke point, and a tool that does not declare itself read-only is treated as a write and gated by default.
What this is
Think inside your AI world — you stay in command
Unlimitless (Unl to friends) holds what you've settled, reads what your tools are showing, and catches what's changed out in the world — and hands your AI whatever bears on the work, the moment it's needed, without you asking. The right thing, in front of the model, unprompted, with you in command of the call. So you keep moving toward what you set out to build, on top of everything you've already decided.
It plugs into Claude, Claude Code, ChatGPT and Cursor as an MCP connector. Quick to connect, in a couple of steps.
Unlimitless is open now to invited Alpha. Apply for the Beta waitlist to come in ahead of the full launch:
Alpha is invite-only · free at launch.