This site is a tool for your agent
AwesomeTechStack registers its own audit, catalogue and monitoring actions as WebMCP tools, so an AI agent in your browser can run them directly instead of clicking through the interface.
What your browser can do
Checking your browser...
WebMCP is a young proposal - a Draft Community Group Report at the W3C Web Machine Learning Community Group, not a standard. We ship it because we would rather be ready early than describe readiness we do not practise.
The tools this site offers
Tools follow the page, exactly as the interface does. Each one below is registered where it makes sense and withdrawn when you leave - so an agent is never holding a tool for a screen that is no longer there.
Everywhere on this site
Registered on every page.
search-technologiesread-onlySearch AwesomeTechStack's catalogue of web technologies by name. Returns matching technologies with their category, homepage and detail page on this site. Reads the catalogue already loaded in this browser: no account, no network request, no cost.
query- Part of a technology name, e.g. "next" or "cloudflare".
limitoptional- How many results to return (1-25, default 10).
get-technologyread-onlyReturn what AwesomeTechStack knows about one technology: its category, description, homepage, source links and - where the health catalogue covers it - the latest known version and whether the release line has reached end of life. No account, no cost.
name- The technology name, e.g. "Nuxt.js" or "jQuery".
audit-websitechanges datacosts one auditneeds sign-inStarts a full audit of one URL and SPENDS ONE AUDIT from the signed-in account's monthly allowance. Requires a signed-in session. Returns immediately with the audit id and its report URL - it does not wait, because an audit takes several minutes. Poll audit-status with the returned id to learn when it is finished.
url- The website to audit, e.g. "example.com" or "https://example.com/pricing".
audit-statusread-onlyneeds sign-inReturn the current state of one audit: its status, the stage it is working on, its progress, and - once finished - its overall score. Costs nothing and starts nothing.
auditId- The audit id returned by audit-website.
get-account-usageread-onlyneeds sign-inReturn the signed-in account's plan and what is left of it this month: audits used and allowed, monitored sites used and allowed, and any prepaid credit balance. Call this before audit-website to avoid spending an allowance that is already exhausted.
No arguments.
On an audit report
Registered while a report is open, and withdrawn when you leave it.
get-current-reportread-onlyReturn the audit report the person is looking at right now, in full and structured: the Awesomeness Score and its band, the technology, up-to-dateness, quality and AI-readiness sub-scores, the AI-readiness grade with each of its checks, every detected technology with its version and version status, vulnerability and end-of-life counts, and the TLS certificate state. Takes no arguments - it is about this page.
No arguments.
get-audit-reportread-onlyneeds sign-inReturn a finished audit report by its id, in the same structure as get-current-report. Reads an audit that belongs to the signed-in account; it neither starts nor re-runs anything.
auditId- The audit id.
watch-current-websitechanges dataneeds sign-inPuts the website of the report on screen under Stack Watch, so it is re-audited on a schedule and its owner is alerted about new vulnerabilities, version drift, certificate expiry and score moves. Changes the account: it adds a monitored site and counts against the plan's monitored-site allowance. Starts no audit and costs no audit.
No arguments.
On the Stack Watch board
Registered on /user/clients.
list-monitorsread-onlyneeds sign-inReturn every site the signed-in account has under Stack Watch: its domain and path, whether it is watched as own or as a competitor, its last score and last scan, when the next scan is due, and whether it is paused.
No arguments.
list-monitor-eventsread-onlyneeds sign-inReturn the Stack Watch event feed - what changed on the monitored sites: new vulnerabilities, version drift, score moves, certificate events. Optionally for one monitor only. Returns a cursor for the next page when the feed continues.
monitorIdoptional- Limit the feed to one monitor. Omit for every monitored site.
cursoroptional- The cursor returned by a previous call, to read the next page.
watch-websitechanges dataneeds sign-inPuts one website under Stack Watch. Changes the account: it adds a monitored site and counts against the plan's monitored-site allowance. Starts no audit and costs no audit.
url- The website to watch, e.g. "example.com" or "example.com/pricing".
typeoptional- Watch it as your own site ("own", the default) or as a competitor ("competitor").
unwatch-websitechanges dataneeds sign-inRemoves one monitored site from Stack Watch, together with its schedule. Takes the monitor id ONLY, never a domain - call list-monitors first, so the person can see which one is about to go.
monitorId- The monitor id from list-monitors.
scan-monitor-nowchanges datacosts one auditneeds sign-inRuns an out-of-schedule audit of one monitored site and SPENDS ONE AUDIT from the monthly allowance. Returns the new audit id immediately; poll audit-status with it.
monitorId- The monitor id from list-monitors.
On the pricing page
Registered on /pricing.
get-plansread-onlyReturn the AwesomeTechStack plans as the pricing page has them: name, price and currency, and the allowances each one carries. The numbers come from the live plan list, never from this page's text.
No arguments.
What an agent cannot do here
This is a decision, not an oversight. Each of these is reachable in the interface by the person signed in - and unreachable through a tool.
An agent moves no money. There is no task that needs it, and no confirmation path in the spec we would trust with a payment.
Key material. A tool call that returns a key puts it into a model's context.
Standing configuration with a delivery target - a rule that outlives the conversation that created it.
Access control.
Identity, and the irreversible.
The person signs in, not their agent.
Needs no explanation.
In the browser, and on the server
WebMCP works inside the page you have open, with the session you are already signed in to. For agents that run without a browser we serve a separate MCP endpoint with API-key authentication - the two are complements, not alternatives.