
Stackcut answers the money questions about software: is this subscription worth it, what's cheaper, which vendor fits what I'm building, and what does my whole stack cost.
Ask it about:
https://stackcut.io/mcp/publicStackcut checks what software costs and what to cut, using public prices it keeps current. Pick the tool by what the user wants: several subscriptions to review or a stack to cut: (it asks for the few facts that matter), then ; the paid services in a codebase: applied on their machine, then ; one product they pay for ('cheaper than X?', 'is X worth it?'): ; picking vendors for something they're building: , or for a whole product brief; a whole workflow change (billing across projects, hosting, email, AI routing): ; browsing by feature or category: (intent 'new' for a job they pay for no tool for yet: costs, not savings); cancelling, pausing or paying less for any subscription (streaming, apps, software): ( to browse by action); negotiating an existing bill or evaluating a retention offer: . After using guidance, call for a missing vendor or route, changed step, source issue, expected human handoff or guide_matched; send only structured fields and an opaque retry key. Unknown language or market stays omitted. After the user reports an action, ties it to the current packet revision and route; it does not verify success. Never send a negotiation message or accept an offer without the user's approval. Pass what the user pays, seats, the features they use and usage numbers (usage: {emails_per_month: 40000, domains: 4, db_gb: 3}, or evidence such as '40,000 emails/mo'). A path is eligible, or a pick fits, only after every check passed; needs_info, fits_if_checks_pass, a candidate path and a hypothetical total mean something is still open, so ask the questions returned instead of recommending the change. Savings are modeled from public list prices, not guaranteed: say so. Ask before any cancel, purchase or sign-up. Send names, numbers and prices only: never source code, secrets, raw invoices or personal data. This endpoint keeps no routine tool arguments or answer contents. It counts calls (tool, success, timing, client name) and which catalog recipes it showed; exit tools also log only result counts, qualification flags and public evidence statuses. Explicit and submissions store their disclosed structured fields, including optional offer terms or a public vendor domain. Feedback and outcome reports remain unverified. For a personal Stackcut plan code (sca_…), connect https://stackcut.io/mcp instead.
None in the store listings or on the server