MCP Server Change and Health Monitor

Give this actor a list of MCP servers and get one row per server: did it answer, how fast, what tools it exposes now, and what changed since the last run. Changes are marked breaking, warning, or info. Run it on a schedule and it compares each run with the last one by itself. You pay per server checked; Mamba Labs pays the platform costs.

Features

• It speaks MCP. The actor connects as an MCP client (initialize, then the list calls), so it sees what an agent sees. You do not need to export a tool catalog.
• Three ways in. HTTP endpoints (streamable HTTP, then the older HTTP plus SSE), npm packages started over stdio, or search terms for the official MCP Registry, in any mix.
• Changes marked by what they break. Breaking: a tool or argument removed, an argument newly required, a type changed, an allowed value removed. Warning: a tool description, annotation, or output schema changed. Info: a tool or optional argument added.
• Description changes are reported. An agent reads a tool description as instructions, so a silent edit to one is reported with the before and after text.
• Health. Answer time, tool list time, server version, last time up, and down since. A server gets one retry before it is reported down.
• Scheduled runs compare themselves. The last result is saved under your Snapshot name in your Apify account. Or pass a SNAPSHOT record from any run.
• Cost guards. A max charge per run, and a cap on servers per registry search.

How the data is read

The actor connects straight to each server from Apify's network. A monitor should see a server the way a client does, so it does not go through a proxy. It uses no login of its own, no other developer's actor, and no key from Mamba Labs. Registry searches read the public API of the official MCP Registry.

Private servers. Put your headers in auth_headers, keyed by site. Each header goes only to its own site and is never sent across a redirect to another site. The field is encrypted by Apify, and the actor never logs it or saves it. Fixtures and the refused path are how this was tested: no real private server was used, so the first buyer's credentials are the live test.

npm packages. A package you list is downloaded and started inside your run, so its code runs there. The actor starts it under an unprivileged user with no environment variables and no install scripts, and refuses to start it if that user can read the run's Apify token. It can still reach the internet. List only packages you trust. Packages found through a registry search are listed in the row but never run.

Input

Field	Type	Default	What it does
server_urls	array		HTTP endpoints of MCP servers. Streamable HTTP is tried first, then the older HTTP plus SSE transport.
npm_packages	array		npm package names, optionally with a version, such as @modelcontextprotocol/server-everything. Each is installed and started inside your run. The code in a package you list runs in your run, so list only packages you trust. No environment variables are passed to it.
registry_search	array		Search terms for the official MCP Registry. The HTTP endpoints of matching servers are checked. Packages found this way are listed but never run, because running code picked by a keyword is not safe.
max_servers	string	25	Most servers to check per registry search term, up to 200. Default 25.
snapshot_key	string	default	Name of the memory that holds the last run's results, so a scheduled run finds changes on its own. Use a different name per list. Default default.
previous_snapshot	string		Optional. The JSON of a SNAPSHOT record from an earlier run, or an https link to that record. It takes the place of the saved snapshot for this run.
timeout_seconds	string	20	How long one request to a server may take before it counts as no answer, 3 to 60. Default 20. A package gets 60 more seconds to install.
include_resources_prompts	string	true	Also list the server's resources and prompts and report what changed in them. Default true.
auth_headers	string		Optional, for servers you are allowed to check. JSON keyed by site, for example {"https://mcp.example.com%22:%7B%22Authorization%22:%22Bearer abc"}}. Each header goes only to its own site. Encrypted by Apify, never logged, never saved in a snapshot.
max_charge_usd	string		Optional. The run stops before a charge that would pass this amount. Free rows are always returned.

Output

Field	Type	What it holds
row_type	string	One of: server_check, registry_query, run_summary. run_summary is the one free row written when the input is refused.
server	string	The URL or npm package checked. On a registry_query row, the search term.
source	string	url, npm, or registry: how this server got on the list.
registry_name	string	The server's name in the official MCP Registry, when it came from a search.
status	string	up (answered and listed its catalog), down (no valid answer after one retry), auth_required (answered 401), blocked (answered 403 or 429), invalid (bad entry), or not_run.
change_status	string	baseline (first time seen, nothing to compare), unchanged, changed, or not_compared (the server gave no catalog).
transport	string	http (streamable HTTP), sse (the older HTTP plus SSE), or stdio.
latency_ms	integer	Time for the server to answer initialize. For a package this includes the install.
list_latency_ms	integer	Time to read the full tool list.
http_status	integer	The HTTP status behind a status other than up, when there was one.
server_name	string	The name the server reports for itself.
server_version	string	The version the server reports for itself.
protocol_version	string	The MCP protocol version the server agreed to.
tool_count	integer	How many tools the server exposes now.
resource_count	integer	How many resources it exposes now. Empty when it has none or the option is off.
prompt_count	integer	How many prompts it exposes now. Empty when it has none or the option is off.
catalog_hash	string	A hash of the whole catalog. It changes when any compared part changes.
tools	array	Every tool now: name, hash, and the first 300 characters of its description.
changes	array	What changed since the last run. Each entry has severity (breaking, warning, or info), kind, item, and detail.
change_count	integer	How many entries are in changes.
breaking_count	integer	How many changes break a caller: a tool or argument removed, an argument newly required, a type changed, an allowed value removed.
previous_checked_at	string	When the snapshot this run compared against was taken.
last_up_at	string	The last time a run found this server up.
down_since	string	When runs first found this server down, while it stays down.
server_count	integer	On a registry_query row: HTTP endpoints from this search that will be checked.
skipped_packages	array	On a registry_query row: matching servers that have only a package. They are listed, not run.
checked_at	string	When this server was checked, UTC.
row_status	string	ok, or why this row carries no data: no_results, no_change, partial, blocked, or error.
error_reason	string	Why the row is not ok. A refused input reads input_invalid: followed by what to fix.
billable_events_json	string	What this row was billed as, for example {"server-check":1}. {} means free.
scraped_at	string	When the row was written, UTC.

Pricing

Pay per event. You pay for a run start and for each server with a definite answer: up and compared, or down after a retry. A package costs more than a URL because each one is a full install. A down result is charged because it is the answer you asked for and the check took the full wait. A server that answers 401 (auth_required), 402, 403, or 429 (blocked), an invalid entry, a package name npm does not have, a package that was not run, a registry row, and a refused input are free.

Event	Free plan	Bronze	Silver	Gold
server-check, one MCP server checked	$0.005	$0.00475	$0.0045	$0.00425
package-check, one npm package installed and checked	$0.02	$0.019	$0.018	$0.017
apify-actor-start, once per run (one per GB of memory; the default 512 MB is one)	$0.002	$0.002	$0.002	$0.002

Per 1,000 rows of server-check: $5.00 on the free plan, $4.25 on Gold.

Free plan cap: 1.50 USD of evaluation a month, about 300 rows. Any paid Apify plan removes the cap, and it resets on the 1st. The run stops before a charge that passes your maximum charge per run in Apify or your max_charge_usd input.

Mamba Labs pays the platform costs of your run, Apify proxy traffic included.

Usage examples

Apify Console or API. A scheduled check of two endpoints and one package:

{
  "server_urls": ["https://mcp.deepwiki.com/mcp", "https://knowledge-mcp.global.api.aws"],
  "npm_packages": ["@modelcontextprotocol/server-everything"],
  "snapshot_key": "my-agents"
}

Registry search.

{ "registry_search": ["github"], "max_servers": "25" }

To act only on real changes, filter rows on change_status equal to changed, or breaking_count above 0.

Clay. Number and on or off inputs are text fields, which is what Clay sends. A list input takes a JSON list, so put the column inside a one item list.

MCP. The Mamba Labs MCP server @mambalabsdev/mcp-server-change-health-monitor exposes this actor to Claude and other MCP clients.

Error handling

Case	What you get
Missing or unreadable input	One free row, row_type run_summary, error_reason starting input_invalid: with what to fix. The run ends successfully and charges nothing but the run start
A server gives no valid answer after one retry	One charged row, status down, error_reason says why
A server answers 401	One free row, status auth_required. Add auth_headers to check it
A server answers 402, 403, or 429	One free row, status blocked
A bad URL or package name in a good list	One free row for that entry, status invalid; the others run
A private or local address	Refused as invalid. The actor checks public servers
A package that needs an environment variable or argument to start	One charged row, status down, because none are passed. A package is not retried
The first run for a server	change_status baseline: nothing to compare yet
The run is close to its timeout	The actor stops starting new work 90 seconds before the timeout and delivers what it has

Limitations

• The first run sets the baseline. Changes appear from the second run.
• It lists what a server exposes. It does not call tools, so it cannot tell whether a tool works.
• A server that shows a different catalog to each visitor, or changes its text per session, shows as changed. Read changes to see what moved.
• A package cannot receive environment variables or arguments in this version, so one that needs a key or a path to start reads as down.
• The older HTTP plus SSE transport was tested on fixtures only; no public server using it answered during the build.
• Up to 500 URLs, 25 packages, and 10 registry terms per run, with up to 200 servers per term. A list of more than 8 MB of snapshot keeps hashes only, so changes name the tool but not the old text.
• Answer time is measured from Apify's network in one place, so it is not a global uptime reading.

Mamba Labs actor suite

Actor	Actor ID
Account Hiring Watch	gRAZA5WBi1r6L0gy7
Agent Accessibility Auditor	anxbRv0lKrpQ1pnua
Agent to Brokerage Resolver	ePRRGLxOGyaVHO4dI
AI Answer Visibility Tracker	RJZvzKniYK4nswcMg
AI Tooling Detector	EwkHhmqiuJgRoVEbE
API Deprecation and Breaking Change Monitor	pHsbS3xXUK8Jqutz5
App Store and Google Play Review Monitor	MaCMEhGrbchxQZ92G
B2B Reddit Conversation Monitor	SoDryzzsjmY3SsiyG
Bluesky Brand Presence Mapper	eLpxzP4IuXznlFVND
Brand Sponsor Finder	PPgK6xrgOmvL7Nrys
Bulk WHOIS and RDAP Domain Age Lookup	MTrkG5FLU6SjqLdJY
Career Change Signal Engine	vX2Yga1a6wGaXaP14
Company Careers Page Jobs, 17 ATS Boards	izP0rfIgf5YaRZfIh
Company Change Event Feed	oX44rS0fkEJ3rXLWe
Company Discovery List Builder	jch0HdaZnpbMqlcGS
Company Firmographic Enricher	YlUtLWjfPpqykmB8g
Company Funding Record	OFS4Mt1gyYNtGfUbE
Company Headcount Estimator	aHC4GROnNt6kfZe5r
Company Identity Resolver	lr8fTRAmZCBZmuwwh
Company Role Email, Phone and Address Finder	4mMncKaJykiq94grz
Company Social Presence Mapper	4k6CCemkgBDz18m2h
Competitor Price, Stock, and Changelog Monitor	yxdgCbUQ4PYQWwRNU
Contact Classifier	0lGSeYJmniXhGANnO
Contractor License Search, CA, TX, WA, OR, LA	WBYK3CPI2kdeq4bP1
CRM Contact Job Change Checker	L8UbEmMJC1XYPSZOb
Domain Deliverability Checker	0tVgxI7A6o9jMlxmc
Domain to LinkedIn URL Resolver	3HtnSaqPHOg1Qg5gx
Ecommerce Platform Profiler	5G2NeYjBbQTjaTn9L
Ecommerce Store Discovery	PeHu4sLyFvQHXUUrX
Email List Verifier, Bring Your Own Key	CzVJ4NlaCKk80xpgE
Event Presence Index	WLhMy8fMDgsxdYxv5
Expert Witness Finder, US and Australia	LIpwlgfE3LaQBaXd4
Funding and Press Signal Scanner	FS13X6dhQVgX3XOM6
GDPR Cookie Banner and Consent Checker	TrKyOS0KBoGC35nbr
GitHub Organization Signal Scanner	ahF8RRpv5mzPao5CN
Google Ads Transparency Center Scraper by Domain	1eOwxhBVxzvOXAm4s
GTM Hiring Signal Scraper	D7O1SA2EqwHGsGr1P
GTM Signals Aggregator	xKdRfnfFNkdMpFuNs
ICP Fit Scorer	W161DT8W4kW55dMFh
Influencer Change Monitor	d2VVgahNL6UmcLkhg
Influencer Lead List Builder	KnmByszcv135yM30G
Influencer Profile Scraper	EqEnklDVMIcB1V499
Influencer Search Across TikTok, Instagram and YouTube	tpkedmloIIWtXx6sg
Influencer Talent Agency Lookup	zCuX4Mgyg6JvXgGzd
Instagram Facebook Brand Presence Mapper	HXh65I0JBjPsunbY8
Integration and Marketplace Presence Mapper	V0guNuvM85Q61f2gj
Job Board Keyword Signal Scanner	4DvqpvhMR74NLcDDY
Job Posting Monitor	QCfICD9kOPiOdj5iI
Journalist and Media List Builder	mAfizsZC8qRoR0ecS
Launch Directory List Builder	cRq3bmvEZxZtioQ7U
Lead File Import Checker	RwlGVXTf46LqZdUgt
Legal Entity Resolver	KHFyPCDIx7CyqULYm
Link in Bio Scraper and Newsletter Detector	OorucdheTIgu7RFzK
LinkedIn Company Page Mapper	zOPm6A9JEhUGcIdFW
LinkedIn Jobs Feed	SH6N0d8cz46BhCC8B
LinkedIn Post Tracker	oiGLNPuaf5BRaz9K5
Local Business Lead Builder	qMN97aoHhsAS5bznV
Local Rank and Map Pack Tracker	gv0kyHGDIg8XscclB
Lookalike Companies from Seed Domains	e2O426WBAbLLQL4Z2
Manufacturer and Supplier Finder	Qcsdmd0lUfPzVkonh
Meta Ad Library Ads by Company, Graph API	J1GWlSXfSGU3u8Wng
Mobile Phone Waterfall	6OOgPktf7RIa5XCOP
New Hire Finder	cKuc38p4nYHNF3IW8
Newly Registered Companies, UK, Norway, Finland, France	g2WfZhWvCtc57g2Jo
News and Media Mention Monitor	gol1PPUKrsIk4QLBd
Newsletter Sponsor Tracker	qSwpqxD1TiPZpLQHv
Open Tender and RFP Finder	kPTGz6t6TFoUUyfaO
Outbound Stack Detector	v43UJC8r7qW7cBSTG
Page Finder and Extractor	TpurgcOZbVnknlaiC
Partner and Member Directory Extractor	ZX0XJPPUvnFJxeC1a
People Finder & Email Verifier	LEgloG7tWgQl4o6TD
Person LinkedIn URL Resolver	ssuyPV8LnEsL8qbnb
Person Social Footprint Mapper	jfQXa2zW9FHCJot9K
Personal Email to Company Resolver	AefT5Ye6YNPFQwJBf
Pinterest Brand Presence Mapper	d1HvjZfKhgvWsoEwn
Podcast Guest and Host Finder	o2MGSvgbSLeeXOsIo
Product Hunt Daily Launches with Maker Domains	O4VX63drWK9Y9x641
Prospect Engine	q9rT6v64nUJdPexOQ
Public Feature Board and Roadmap Monitor	rcHF5gj5r2Vqnry35
Publishing Frequency Tracker	TbLwaUUATdYb6wp4N
Recent Funding Rounds Feed	KueVp9MOz7iOoAgDp
Reddit Brand Presence Mapper	pKQZmDaT70SEb54bZ
Reddit Search and Thread Scraper	4Gsth5cP2Xt9V2xDN
Reporting Window Finder	uINxR7a1IW8qUTRUX
Sequencer Event Poller	S6s8Rjd61tGdrMRk4
Sequencer Lead Push	0Jv27VeWM5tSZQs9x
Shopify App Store and Chrome Web Store Listings	gaY0BzkLyN77mD3wI
Substack Newsletter and Posts Scraper	UEFsdwiPLmPNVy9xd
Team Page People Extractor	j8a9zX0k4mNbvDbWK
Tech Stack Signal Detector	qyd7nNyqFPelQViBx
Threads Posts by Keyword, Exact Match	2OT8K8RiEEWtaj2MW
TikTok Brand Presence Mapper	fco8FvoJcSafbeyd1
Trustpilot Reputation Enricher	3lIqfCt61j4hqE1xz
US and UK Contract Award Winners by Company	zhEtllASykOcx9hJ8
VC and Accelerator Portfolio List Builder	chH8hrfuoqqolW06m
Website Accessibility Checker, WCAG 2.2 AA	P3jBj3FRghKEhX52z
Website Phrase and Business Model Classifier	CBuZ9k7sRwBeiKiCS
Website Traffic Rank Estimator	GnOFUzlNVIoXgeJjH
Work Email Waterfall	OT6xqTFThC0Rjf7Zj
Workplace Program Detector	Li6pPDhqbdzFz3h7G
X Brand Presence Mapper	oLOgadhHUkNAyDA3w
YouTube Channel from Company Domain	6hSpD3kd2s2IQ4VSs

Built by Mamba Labs.