Connect inetGeek to an AI assistant
Everything on this site — sourced pricing, comparisons, stack recommendations, live DNS tools — is also reachable as an MCP server, so an assistant like Claude or ChatGPT can read it directly instead of guessing from training data that may predate this week's price change.
What it gives an assistant
Twenty tools. All but one are read-only: send_feedbackrecords a note for the person who maintains the site and can change nothing. Every fact the others return is the same sourced, dated figure published on the matching page here — this is a distribution layer for data already verified, not a place a new claim originates.
list_categories
The categories this site tracks, each with its current provider count.
list_providers
Every provider in one category with its free tier, entry price and the editorial "best for" line — the orientation call before a per-provider lookup.
filter_providers
Providers that document every fact key you name (e.g. SOC 2 and HIPAA), each with the value, source and verification date.
search_infrastructure
Full-text search over every published page — providers, comparisons, DNS records, tools and stacks — including the fact labels each provider documents.
get_provider
The full sourced fact sheet for one provider — pricing, limits, regions and startup-credit programs, each with a source and a verification date.
list_startup_credit_providers
Every provider documented as offering a startup credit program, with the amount and eligibility — a filter, not a per-provider lookup.
get_provider_score
One provider's iScore with its interval and rank, every pillar and criterion with the fact behind it, and the criteria its peers document that it does not — what a provider would publish to move the number.
list_alternatives
Every peer in a provider's category ranked by iScore, with score, interval, entry price and where each differs from the provider named.
compare_providers
The sourced side-by-side comparison for a curated pair, where one exists.
list_comparisons
Every published comparison, optionally only those involving one provider — so an agent knows a pair exists before asking for it.
get_stack_recommendation
Every host that documents real support for a framework or runtime, ordered by strength of documentation.
plan_infrastructure
The /plan/ planner: one suggested provider per layer for a stage or a hand-picked set of layers, a floor-price total, and the layers left out of it for having no clean starting price. Takes the project’s context — what it already uses, the platform it runs on, what it excludes — so a tie is broken by the project and a covered layer is left alone; ties are reported as ties.
estimate_monthly_cost
What each provider that publishes a per-volume price would charge a month at a stated volume (web analytics, email and auth): plan fee plus overage, cheapest first, with the vendor page and read date. inetGeek’s arithmetic on published rates, not a quote.
get_ledger
Every vendor page read and found silent on a criterion, with the page and the date; every criterion a product cannot have; every case where two pages of one vendor disagree, with both quotes; every changed figure. A silence record is an observation, not a claim the provider lacks the thing.
get_changelog
Every figure that moved — before, after, source and date — optionally since a date or for one provider.
get_dns_record
The reference page for one DNS record type or email-authentication mechanism, as text.
check_dns
A live DNS lookup for a domain's A/AAAA/CNAME/MX/TXT/NS records.
check_spf
A live SPF evaluation, including the ten-lookup limit.
check_dns_propagation
A live check of one record across four independent public resolvers.
send_feedback
Report a wrong figure, a broken or confusing tool, or something missing to the person who maintains the site. The only tool that writes: it stores a note and changes nothing; a wrong-figure report is checked against the provider's own page first.
What it won't do
compare_providers only answers for the curated pairs already published here — an uncurated pair returns a plain "not published yet," not a guess assembled on the spot. Cost at your volume is the one place it does arithmetic: estimate_monthly_cost applies a vendor's published per-volume rates to a monthly figure you give it (web analytics, email and auth so far) and says beside every number that it is our arithmetic, not a quote. A provider with no readable per-volume price is named, never guessed at, and no tool projects a cost for anything else: this site records what a provider documents, not an estimate of what heavy traffic would cost you, because a vendor's own spend estimate is not a price either — see the methodology. An assistant can reason about scaling readiness from the real limits get_provider returns — connection counts, egress pricing, regional coverage — but any dollar figure beyond those rows is its arithmetic, not a number this site asserts.
iScore, for two kinds of reader
Two tools carry iScore, the site's own reading of a provider against its category peers: a number from 0 to 100 with an 80% interval, computed at build from the same sourced facts by rules published on that page. Both return the interval with the number and mark a rank whose interval overlaps the next one as an ordering, not a finding — an assistant that drops either has dropped the honest part.
Choosing a substitute. list_alternatives takes a category id and a provider and returns every peer ranked by iScore — score, interval, free tier, entry price, the editorial line, and up to two sourced facts where each differs from the one named. It is the same list as the provider's alternatives page. Follow a row with get_provider_score for the breakdown, or compare_providers where a curated pair exists.
Reading your own score. get_provider_score returns every pillar with its weight and interval, every scored criterion with the fact it was read from and the peer median it was measured against, and the criteria scored for peers but not for you — undocumented, or stated without a number or a yes/limited/no. That last list is the answer to "what would move it": publish those facts plainly on a page a fetch can read. The score follows the facts and is never edited directly; a wrong figure goes through corrections.
Both take a category id — database, paas, llm-api — where get_provider takes a URL section; a miss returns the valid list. get_provider also carries the iScore section from the provider's markdown mirror, so an assistant that already has the fact sheet has the number too.
Connect it
No account and no API key — the server is open and unauthenticated, because everything it answers with is already public on this site. The URL is the same for every client:
https://inetgeek-mcp.palashbagchi.workers.dev/mcp
ChatGPT
Settings → Connectors → Add custom connector → paste the URL above → no authentication.
Claude.ai
Settings → Connectors → Add custom connector → paste the URL above → no authentication.
Claude Code
claude mcp add --transport http inetgeek https://inetgeek-mcp.palashbagchi.workers.dev/mcp
Cursor and other MCP clients
Add to the client's MCP config file:
{
"mcpServers": {
"inetgeek": { "url": "https://inetgeek-mcp.palashbagchi.workers.dev/mcp" }
}
}Step-by-step setup for each client, example prompts that were run against the live server, how to give it your project's context, and a tool reference generated from the server itself are in the docs.
What the server keeps
Nothing you ask it. The DNS, SPF and propagation tools query public resolvers live and return the answer; the domain you looked up is not written down, and neither is any other argument you pass — not the provider you asked about, not the stack you were planning.
The one exception is a report an agent chooses to send withsend_feedback: a wrong figure, a broken tool, something missing. That is stored so a person can read it — what the agent wrote, the time, and the client's User-Agent, never the IP address — and deleted 90 days after it is closed, or a year after it arrives. A wrong-figure report is checked against the provider's own page before anything changes, exactly as an emailed correction is. Theprivacy policy has the detail.
Three things are counted, so that whether this server is used at all stops being a guess: how many times a client completes a handshake, along with the product name it declares for itself; how many times each tool is called, by tool name; and the User-Agent, which names the software making the call. No IP address, no other header, no argument value, and nothing that identifies a person.
The User-Agent is recorded because without it most calls are anonymous, and that is worth saying plainly rather than leaving as a technicality. A client names itself once, in the handshake. This server then holds no session and issues no key, so every call after that carries no identity at all — on the day this counter shipped, two clients were connected and actively calling tools while the handshake figure read one. The User-Agent closes that gap, and it names a piece of software rather than a person.
None of it counts installs, and no arrangement of it could. Because there is no session and no key, a client reconnecting after a restart is indistinguishable from one arriving for the first time. Read these numbers as activity.
See the privacy page for what this site collects generally, and report a problem → if a tool answers something wrong.
Read it, or run your own
The server is open source. Every tool description above is in that repository verbatim — so if you want to know exactly what an assistant is told about a tool before it calls it, the answer is one file, not a guess.
It is stateless: it holds no database and reads this site's public JSON feeds and Markdown mirrors at request time. A copy you deploy yourself therefore serves the same figures from the same origin, which is the point of publishing it — a different tool surface, your own descriptions, or your own Cloudflare account in the path, without a second copy of the data going stale somewhere.
The code is MIT. The data it serves is not part of that licence: it lives here, and each figure stays attributed to the vendor page it was read from.