MCP servers › dev.shveik › agdata
MCP server · registry snapshot 2026-10-23 · last tested 2026-10-23
dev.shveik/agdata
Unblocking and fresh web data for agents: URL to Markdown, YouTube, Maps, Amazon, jobs. Pay per call — the publisher's description
Summary
dev.shveik/agdata is a remote MCP server listed in the official MCP registry at version 1.0.0. On 2026-10-23 ProtocolProbe connected to it in read-only mode and read its inventory: 31 tools, 4 resources and 0 prompts.
Record checks
1 of 2 record checks met · 7 do not apply
- met: Registry entry is active and says how to run it
- not met: Source repository link opens
- not applicable: Repository is not archived
- not applicable: Code pushed to the repository in the 90 days before our check
- not applicable: Repository publishes a security policy (SECURITY.md)
- not applicable: Package names its licence with a standard (SPDX) identifier
- not applicable: The version the registry lists can be installed and is not withdrawn
- not applicable: npm package carries build provenance (attestation)
- not applicable: No known critical or high advisory in the package or its dependencies
- Not ranked by downloads: a remote server publishes no download figure
- Not ranked by stars: no public GitHub repository
From public records, not a safety rating. Each check with its reason · Is it safe to install?
Key facts
- Runs
- remote (hosted by the publisher)
- Transport
- streamable HTTP
- Sign-in
- none needed to open a session
- Tools
- 31
- Described as changing data
- 1 of 31 toolsAI-classified
- Mainly works on
- accounts and data in outside web servicesAI-classified
- Topic
- search-and-web-dataAI-classified
- Pricing
- usage-basedAI-classified
- Last tested
- 2026-10-23
Worth knowing before you connect it
- 1 tool is described as changing something (feedback): our AI model's reading of the descriptions, not something we tested, and the server opened a session without any credential. We did not call it.
Each line comes from a section below, with its source and date.
How fresh this is
- Live-tested
- yes, a real session (not a text-only review)
- Last tested
- 2026-10-23
- Mode
- read-only (no tool calls)
- Latest version
- 1.0.0, published 2026-10-05 (registry, as of the 2026-10-23 snapshot)
- First listed version
- 2026-10-05 (earliest version still in the registry; older removed versions would be earlier)
- Registry data
- snapshot 2026-10-23
- Package and repository
- checked 2026-10-22
- Engine
- mcp-runtime-engine@0.1.0 · sdk@1.32.0
Changes since the last refresh
New in the registry: not in the list of 2026-10-04, listed on 2026-10-23 with version 1.0.0 (published 2026-10-05).
Source: the official MCP Registry list on 2026-10-04 and on 2026-10-23.
What it is
What does this server offer?
Observed inventoryWhat it mainly works on, as our AI model read its tool descriptions: accounts and data in outside web services.
1 of 31 tools are described as changing something (creating, editing, deleting, sending or running).
“Unblocker for AI agents: send {url} and get the page back as clean Markdown, with title, description, language, final URL and HTTP status.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
2 inputs, 1 required
formatstring“how the page comes back in content”urlstringrequired“absolute http or https URL of a public page, e.g.”
- instagramreads or searches
“Instagram posts as JSON.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
3 inputs, 2 required
limitinteger“number of posts wanted, 1 to 20 (default 5).”modestringrequired“profile=account's posts, hashtag=posts under a tag, search=keyword→posts, user=posts from matching users”querystringrequired“username (profile), hashtag without #, or search keywords”
- google-mapsreads or searches
“Google Maps places for a search term, optionally narrowed to an area: {"query":"coffee shops","location":"Prague","limit":5}.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
3 inputs, 1 required
limitinteger“number of places wanted, 1 to 20 (default 5).”locationstring“optional area to narrow the search, e.g.”querystringrequired“Google Maps search term, e.g.”
“Get TikTok videos for a username, a hashtag or a search keyword.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
4 inputs, 2 required
limitinteger“number of videos wanted, 1 to 20 (default 5).”modestringrequired“profile=user's videos, hashtag=videos under a tag, search=keyword→videos”querystringrequired“username (profile), hashtag without #, or search keywords”rawboolean“true returns the provider's own records instead of the compact shape”
- linkedinreads or searches
“Look up one public LinkedIn profile per call.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
1 input, 1 required
querystringrequired“LinkedIn username (e.g.”
“YouTube video search by keyword, or the videos of one channel.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
3 inputs, 2 required
limitinteger“number of videos wanted, 1 to 20 (default 5).”modestringrequired“profile=channel's videos, search=keyword→videos”querystringrequired“channel handle without @ (profile) or search keywords”
- twitterreads or searches
“Tweets from X (Twitter) for one account or a search query.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
3 inputs, 2 required
limitinteger“number of posts wanted, 1 to 20 (default 5).”modestringrequired“profile=account's tweets, search=keyword→tweets”querystringrequired“handle without @ (profile) or search keywords”
- redditreads or searches
“Search Reddit or read a subreddit.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
4 inputs, 2 required
limitinteger“number of posts wanted, 1 to 20 (default 5).”modestringrequired“profile=user's posts, hashtag=subreddit's posts, search=keyword→posts”querystringrequired“username (profile), subreddit without r/ (hashtag), or search keywords”rawboolean“true returns the provider's own records instead of the compact shape”
“Keyword search on Amazon.com (US store): up to 20 products per call, each filled in from its product page.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
2 inputs, 1 required
limitinteger“number of products wanted, 1 to 20 (default 5).”querystringrequired“Amazon search term, e.g.”
- google-searchreads or searches
“Google Search results (SERP) for a query, as JSON.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
2 inputs, 1 required
limitinteger“number of results (billed in pages of about 10) wanted, 1 to 20 (default 5).”querystringrequired“Google Search query, e.g.”
- google-maps-reviewsreads or searches
“Google Maps reviews API for AI agents: send a place_id or a Google Maps place URL and get up to 50 reviews as JSON: stars, text and its translation, date, likes, the owner's reply, plus the place's n…”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
4 inputs, 1 required
languagestring“language the reviews are translated into (textTranslated)”limitinteger“number of reviews wanted, 1 to 50 (default 10).”placestringrequired“a Google place_id (27 characters starting with ChIJ or GhIJ), or a Google Maps place URL: https://www.google.com/maps/place/...”sortstring“order of the reviews”
“YouTube transcript API for AI agents: send a video URL or 11-character id and get its published subtitles as timed JSON segments, plain text, SRT or WebVTT, with title, channel, duration, views and p…”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
3 inputs, 1 required
formatstring“json: timed segments in transcript; text, srt or vtt: the transcript as one string”langstring“subtitle language; any takes English if the video has it, else the first track”videostringrequired“one video: an 11-character video id, or a watch, shorts, live, embed or youtu.be URL of one video (not a channel, playlist or search)”
Show 19 more tools
“Screenshot API for AI agents: send {url}, get JPEG, PNG, WebP or PDF as base64 JSON.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
10 inputs, 1 required
colorSchemestring“emulated prefers-color-scheme”delayMsinteger“extra wait after the page loaded, in milliseconds”devicestring“viewport: desktop 1920x1080, desktop_hd 1366x768, mobile (phone), tablet”formatstring“image or document format of the capture; pdf is always the full page”fullPageboolean“capture the whole scrollable page (height is capped at 16,384 px), not just the viewport”heightinteger“custom viewport height in pixels (ignored with fullPage)”selectorstring“CSS selector: capture only the first matching element”urlstringrequired“absolute http or https URL of a public page, e.g.”waitForSelectorstring“CSS selector to wait for before capturing”widthinteger“custom viewport width in pixels (overrides the device)”
- linkedin-jobsreads or searches
“LinkedIn jobs search API for AI agents: send a job title or keywords and an optional location and get up to 50 public job listings as JSON: title, company, location, posting date, employment type, se…”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
4 inputs, 1 required
limitinteger“number of jobs wanted, 1 to 50 (default 10).”locationstring“city, region or country, e.g.”postedstring“only jobs posted within the last day (24h), week or month (default: any time)”querystringrequired“job title, skill or company, e.g.”
- indeed-jobsreads or searches
“Indeed jobs search API for AI agents: send a job title and an optional location and country and get up to 50 job listings as JSON: title, employer, location, posting date, job types, benefits, salary…”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
5 inputs, 1 required
countrystring“two-letter country code of the Indeed site searched (uk for the United Kingdom)”limitinteger“number of jobs wanted, 1 to 50 (default 10).”locationstring“city, state, postal code or "remote", e.g.”postedstring“only jobs posted within the last day (24h), 3 days (3d), week or 14 days (14d) (default: any time)”querystringrequired“job title, keywords or company, e.g.”
- shopify-productsreads or searches
“Shopify store products API for AI agents: send a store's domain and get up to 250 of its public products as JSON: title, vendor, product type, price range, compare-at price, sale and availability fla…”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
2 inputs, 1 required
limitinteger“number of products wanted, 1 to 250 (default 20).”storestringrequired“the store's domain, e.g.”
- google-newsreads or searches
“Google News search API for AI agents: send a keyword or phrase and get up to 30 current news articles as JSON: headline, publisher, publish time, a short description and the article's link.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
4 inputs, 1 required
countrystring“two-letter country code of the news edition, e.g.”languagestring“two-letter language code of the news edition, e.g.”limitinteger“number of articles wanted, 1 to 30 (default 10).”querystringrequired“keyword or phrase to search Google News for, e.g.”
- amazon-reviewsreads or searches
“Amazon product reviews API for AI agents: send an ASIN or an Amazon product URL and get up to 50 customer reviews as JSON: rating, title, text, date, verified-purchase flag and helpful votes.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
5 inputs, 1 required
countrystring“Amazon marketplace of an ASIN (a URL sets its own; a different value is refused)”limitinteger“number of reviews wanted, 1 to 50 (default 10).”productstringrequired“an ASIN (10 characters, e.g.”sortstring“order of the reviews”starsstring“keep only reviews with this many stars (5 to 1), the positive ones (4 and 5) or the critical ones (1 to 3)”
- google-play-reviewsreads or searches
“Google Play app reviews API for AI agents: send an Android package name or a Google Play app URL and get up to 200 reviews as JSON: rating, text, date, app version, language and helpful votes.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
4 inputs, 1 required
appstringrequired“an Android package name (e.g.”limitinteger“number of reviews wanted, 1 to 200 (default 20).”ratinginteger“keep only reviews with this many stars (1 to 5); omit for all”sortstring“order of the reviews”
- linkedin-companyreads or searches
“LinkedIn company enrichment API for AI agents: send up to 10 LinkedIn company page URLs and get one JSON record per company: name, website, employee and follower counts, founding year, tagline, descr…”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
1 input, 1 required
companiesarrayrequired“LinkedIn company page URLs, 1 to 10, e.g.”
“Web search with page content for AI agents: send keywords and get the top results with each page read as clean Markdown, in one call: title, URL, a short description and the page's text.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
2 inputs, 1 required
limitinteger“number of results wanted, 1 to 5 (default 3).”querystringrequired“search keywords, e.g.”
- x-repliesreads or searches
“Replies to a post on X (Twitter) for AI agents: send the post's URL and get the replies as JSON: text, time, likes, reposts, replies, views and the author's handle.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
2 inputs, 1 required
limitinteger“number of replies wanted, 1 to 50 (default 10).”urlstringrequired“URL of the post on x.com or twitter.com, e.g.”
“YouTube comments API for AI agents: send the URL of a video and get its comments as JSON: text, time, likes and reply count.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
3 inputs, 1 required
limitinteger“number of comments wanted, 1 to 100 (default 20).”sortstring“top comments first (default) or the newest first”urlstringrequired“URL of a YouTube video, e.g.”
- tiktok-commentsreads or searches
“TikTok comments API for AI agents: send the URL of a video and get its comments as JSON: text, time, likes and reply count.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
2 inputs, 1 required
limitinteger“number of comments wanted, 1 to 100 (default 20).”urlstringrequired“URL of a TikTok video, e.g.”
- instagram-commentsreads or searches
“Instagram comments API for AI agents: send the URL of a post or reel and get its comments as JSON: text, time, likes and reply count.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
2 inputs, 1 required
limitinteger“number of comments wanted, 1 to 100 (default 20).”urlstringrequired“URL of an Instagram post or reel, e.g.”
- ebay-soldreads or searches
“eBay sold listings API for AI agents: send a keyword, get sold items as JSON for resale comps: sold price, currency, shipping, total, sale date, condition, listing type, bids, seller feedback.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
5 inputs, 1 required
conditionstring“only new or only used items (default any)”daysinteger“sold within the last this many days (1 to 90, default 30)”keywordstringrequired“what was sold, as typed into eBay's search, e.g.”limitinteger“number of sold listings wanted, 1 to 100 (default 20).”sitestring“the eBay marketplace searched”
- meta-adsreads or searches
“Meta ad library API for AI agents, for competitor research: send a keyword or an advertiser (name or page id) and get its public ads as JSON: advertiser, start and end date, whether it is still runni…”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
5 inputs
activeboolean“true: only ads running now; false: active and inactive ads (default true)”countrystring“two-letter country code the ads are shown in, or ALL (default US)”keywordstring“search the ads by this word or phrase (give this or page, not both)”limitinteger“number of ads wanted, 1 to 100 (default 20).”pagestring“the advertiser: a page name or its numeric page id (give this or keyword, not both)”
- linkedin-profile-emailreads or searches
“LinkedIn profile with a verified work email for AI agents: send one LinkedIn profile URL and get the public profile (headline, about, location, current and past positions, education, skills) plus the…”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
1 input, 1 required
urlstringrequired“LinkedIn profile page URL, e.g.”
- linkedin-postsreads or searches
“LinkedIn posts API for AI agents: send one LinkedIn profile or company page URL and get its recent posts as JSON, each with text, date, likes, comments, shares, author, post URL and the reposter when…”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
2 inputs, 1 required
limitinteger“number of posts wanted, 1 to 20 (default 5).”urlstringrequired“LinkedIn profile URL (https://www.linkedin.com/in/<name>/) or company page URL (https://www.linkedin.com/company/<name>/); one URL per call”
- search-plusreads or searches
“Search API for AI agents: web or news search with up to 30 results per call, each with a title, URL and a query-relevant excerpt.”
Publisher declares: read-only yes · destructive no · idempotent no · open-world yes
6 inputs, 1 required
categorystring“restrict web results: developer (repos, issues, docs), github or pdf”exclude_domainsarray“leave out these hostnames; not together with include_domains”include_domainsarray“only results from these hostnames, e.g.”limitinteger“number of results (billed in pages of 10) wanted, 1 to 30 (default 10).”querystringrequired“search keywords, e.g.”typestring“web (the default) or news”
“Free.”
Publisher declares: read-only no · destructive no · idempotent no · open-world no
4 inputs, 1 required
contactstring“optional: how we can reach you or your operator”kindstringmessagestringrequired“what worked, what did not, what you miss; no secrets or private data”routestring“the route the feedback is about, if any”
Each quoted sentence is the first sentence of the publisher’s own description. The kind of tool and the “changes data” mark are our AI model’s reading of that description, not something we tested: two differently worded runs had to agree, otherwise no label is shown, and nobody has checked them by hand. A tool without the “changes data” mark is described as only reading, searching or calculating, or the two runs did not agree. No tool was called. Read the publisher’s full description before relying on it. “Publisher declares” lists the hints the server itself sends with the tool (MCP annotations: read-only, destructive, idempotent, open-world); they are the publisher's claims, not checked by us. Inputs are the parameters in the tool's own input schema: name, declared type, whether it is required, and the first sentence of its description. Read 2026-10-23.
How well are its tools described?
Features #4 and #5
How clearly this server's own tool descriptions say what a tool does, and how consistent its tool list is. The first part is an AI model's yes/no reading of each description; the second part is counted. We do not turn it into a grade.
Descriptions
AI reading of each description
- Says what the tool does
- 100% of 31 tools
- Says what it returns
- 97% of 31 tools
- States limits, defaults or errors
- 96% of 27 tools
- Says when to use it
- 100% of 31 tools
- Items covered per tool
- 3.9 of 4 on average
Counted from the listing
- Parameters with a description
- 99% of 103
- Tools with a description
- 100% of 31
- Typical description length
- 364 characters
The tool set
Counted from the listing
- Tools
- 31
- Size of the list
- about 9,637 tokens for an agent to load (characters ÷ 4)
- Naming
- 21 of 31 names use kebab-case
AI reading: easy to mix up
No pair found.
The AI items cover 31 of 31 tools: for a tool where two differently worded runs disagreed, that item is left out. They were not checked by hand, and no tool was called. “Easy to mix up” means two tools whose names or descriptions overlap without saying how they differ; the AI found none. We do not rate whether the tool set is complete or whether side effects are spelled out, because our AI reading of those was not reliable enough.
How do I add it to my client?
Config from the registry entry
From the remote endpoint (Streamable HTTP)
We connected to this URL on 2026-10-23 (a read-only session: no tool was called).
Add to CursorAn install link in the format Cursor documents, holding the Cursor config below.
Claude Desktop or claude.ai
Settings → Connectors → Add custom connector, then paste: https://agdata.shveik.dev/mcp
Cursor · .cursor/mcp.json in a project, or ~/.cursor/mcp.json
{
"mcpServers": {
"agdata": {
"url": "https://agdata.shveik.dev/mcp"
}
}
}VS Code · .vscode/mcp.json
{
"servers": {
"agdata": {
"type": "http",
"url": "https://agdata.shveik.dev/mcp"
}
}
}VS Code (command line) · Bash or zsh terminal (macOS, Linux)
code --add-mcp '{"name":"agdata","type":"http","url":"https://agdata.shveik.dev/mcp"}'Claude Code
claude mcp add --transport http agdata https://agdata.shveik.dev/mcp
Try it in the MCP Inspector · runs on your machine, opens in your browser · Node 22.19+
npx @modelcontextprotocol/inspector@2.11.0 --server-url https://agdata.shveik.dev/mcp --transport http
- The server runs on the operator's infrastructure; it may ask you to sign in when you first connect.
- The MCP Inspector lists the server's tools, resources and prompts; a tool you call there runs for real.
Built from the server's registry entry with each client's documented format (checked 2026-10-14; the Cursor link and the VS Code command 2026-10-19; the MCP Inspector command 2026-10-21). Values in < > are yours to fill in; this page never contains a key. Where the server needs a setting the entry does not declare, it can still fail to start.
Does the registry listing match what we found?
Feature #6The registry verifies a package once, when it is published, and does not re-check it. A difference below is about the listing, not a judgement of the publisher.
| Registry says | We found | Result | Source · date |
|---|---|---|---|
| Remote endpoint agdata.shveik.dev/mcp | Session opened | matches | our test · 2026-10-23 |
| No credential declared as required | Opened a session without any credential | matches | our test · 2026-10-23 |
Choosing
Record checks
Public records · not a safety rating1 of 2 record checks met · 7 do not apply
- Not ranked by downloads: a remote server publishes no download figure
- Not ranked by stars: no public GitHub repository
- MetRegistry entry is active and says how to run itActive, with a package or a remote address.
- Not metSource repository link opensThe registry entry gives no repository link.
- Does not applyRepository is not archivedNo public GitHub repository to read.
- Does not applyCode pushed to the repository in the 90 days before our checkNo public GitHub repository to read.
- Does not applyRepository publishes a security policy (SECURITY.md)No public GitHub repository to read.
- Does not applyPackage names its licence with a standard (SPDX) identifierRemote server: it ships no package.
- Does not applyThe version the registry lists can be installed and is not withdrawnRemote server: it ships no package.
- Does not applynpm package carries build provenance (attestation)Remote server: it ships no package.
- Does not applyNo known critical or high advisory in the package or its dependenciesRemote server: it ships no package.
Read from public records only: the registry entry, npm/PyPI, GitHub and advisory databases, checked 2026-10-22. No tool was called. Only “met” and “not met” are counted, so the total differs between servers; a check that does not apply or that we could not run is listed, never counted against it. It is not a safety rating. How each check works · Leaderboards
Common questions about dev.shveik/agdata
Answered from the data aboveIs dev.shveik/agdata safe to install?
We do not rate servers as safe or unsafe. 1 of its 2 applicable record checks are met (facts from public records, each listed on this page). What the records show: 1 of its tools describes changing data (our AI model's reading of the descriptions). These are facts with their sources on this page, not a safety guarantee, and we did not run its tools.
Is dev.shveik/agdata free to use?
Its website (agdata.shveik.dev/) describes a usage-based pricing model. This is an AI-classified reading of the page on 2026-10-23, so check the page itself before relying on it.
Does dev.shveik/agdata need an API key or sign-in?
It opened a session without any credential on 2026-10-23. The registry entry declares no required credential; no declaration is not a guarantee.
Has dev.shveik/agdata been security tested?
Partly. 3 of the 14 checks that apply to it were exercised. We connected and read what the server declares, but did not call its tools: ProtocolProbe calls no server's tools for anything it publishes. This is our choice, not a block by the server. No security verdict is published before verification.
What can dev.shveik/agdata do?
It lists 31 tools. In our AI model's reading of their descriptions (two runs had to agree, nobody checked it by hand), they include tools that read or search. We did not call any tool, so this describes the text, not the behaviour.
Can dev.shveik/agdata change or delete data?
1 of its tools says in its description that it creates, edits, deletes, sends or runs something (our AI model's reading). We did not call any tool, so we cannot say what actually happens.
Found this page useful?
Anonymous · no sign-inA quick, anonymous reaction — no account, no email. One per kind from your browser. The tally updates when the page is next refreshed, and it is only a rough count, never a safety rating or a measured number of people.
Similar servers
Same topic · AI-read
- com.browser-use/browser-usealso has tools that reads or searchescompare with this one
- io.github.upstash/context7Context7also has tools that reads or searchescompare with this one
- io.github.bytedance/mcp-server-browseralso has tools that reads or searchescompare with this one
- io.github.screenpipe/screenpipe-mcpscreenpipealso has tools that reads or searchescompare with this one
- io.github.KKKKhazix/aihotAIHOTalso has tools that reads or searchescompare with this one
- io.github.firecrawl/firecrawl-mcp-serverFirecrawl MCP Serveralso has tools that reads or searchescompare with this one
Active servers whose topic is also “search-and-web-data” and that have at least one tool of the same kind; most kinds in common first, then most GitHub stars. The topic and the tool kinds are an AI model's reading of the publishers' descriptions, not checked by hand. Similar is not a recommendation.
Compare with another server
Side by side: registry entry, our read-only session, the tools and the 18 classification fields. Same values as each page; no score.
Security testing
What did we test, and what did we leave out?
18 checks · features #1, #2
Not probed (we call no server's tools)
We connected and read what the server declares, but did not call its tools: ProtocolProbe calls no server's tools for anything it publishes. This is our choice, not a block by the server. No security score is given; a check we did not run is never counted as passed.
- Inventory: exercised
- Hidden instructions: exercised
- Real execution: not run · our policy
- Secret leak: not run · nothing to check it against
- Internal addresses: not run · our policy
- Unconfirmed changes: not run · our policy
- Bad input: not run · our policy
- Model requests: not run · nothing to check it against
- Sensitive questions: not run · nothing to check it against
- Folder boundary: not run · our policy
- Log leaks: not run · nothing to check it against
- Hidden content: exercised
- Fake assistant turns: not applicable
- Token passthrough: not applicable
- Token audience: not applicable
- Sign-in redirect: not run · our policy
- Consent handling: not testable from outside
- Sign-in metadata: not applicable
3 exercised11 not run (reason given)4 not applicable
Each check ends in exactly one state: exercised, not run (with the reason), or not applicable (with the reason). 3 of the 14 checks that apply to it were exercised. We do not publish security-check outcomes for named servers.
- 1Inventory
tool-listReads the declared tools, resources and prompts.exercised - 2Hidden instructions
description-poisoningLooks for hidden instructions in tool descriptions.exercised - 12Hidden content
audience-hidingChecks resources that hide content from the user.exercised
- 3Real execution
real-executionCalls a tool and checks what really happens.This check calls the server's tools. ProtocolProbe calls no server's tools for anything it publishes; an operator can run this check on their own server with our self-test command.not run · our policy - 4Secret leak
canary-leakPlants a fake secret and checks whether it leaks.This check needs at least one successful tool call to have something to check, and no tool was called.not run · nothing to check it against - 5Internal addresses
ssrfChecks whether a tool can be steered to internal addresses.This check calls the server's tools. ProtocolProbe calls no server's tools for anything it publishes; an operator can run this check on their own server with our self-test command.not run · our policy - 6Unconfirmed changes
write-without-confirmationChecks whether tools that change data act without asking.This check calls the server's tools. ProtocolProbe calls no server's tools for anything it publishes; an operator can run this check on their own server with our self-test command.not run · our policy - 7Bad input
error-handlingSends bad input and watches how the server answers.This check calls the server's tools. ProtocolProbe calls no server's tools for anything it publishes; an operator can run this check on their own server with our self-test command.not run · our policy - 8Model requests
samplingChecks whether the server asks the client's model to act for it.This check needs at least one successful tool call to have something to check, and no tool was called.not run · nothing to check it against - 9Sensitive questions
elicitation-sensitive-dataChecks whether the server asks users for sensitive data.This check needs at least one successful tool call to have something to check, and no tool was called.not run · nothing to check it against - 10Folder boundary
root-boundaryChecks whether the server reaches outside the folders it was given.This check calls the server's tools. ProtocolProbe calls no server's tools for anything it publishes; an operator can run this check on their own server with our self-test command.not run · our policy - 11Log leaks
logging-side-channelChecks whether log messages carry data they should not.This check needs at least one successful tool call to have something to check, and no tool was called.not run · nothing to check it against - 16Sign-in redirect
open-redirectChecks the sign-in redirect for abuse.Testing the sign-in redirect means sending probes to the operator's sign-in service, which we do not do.not run · our policy - 17Consent handling
confused-deputyChecks consent handling when one service acts for another.Depends on the operator's own consent screen, which differs per server and cannot be tested from outside.not testable from outside
- 13Fake assistant turns
fake-assistant-turnChecks prompts that pose as the assistant's own words.The server exposes no prompts.not applicable - 14Token passthrough
token-passthroughChecks whether the server forwards a user's token to other services.The server opened a session without sign-in, so there is no token flow to check.not applicable - 15Token audience
token-audience-validationChecks whether tokens meant for another service are accepted.The server opened a session without sign-in, so there is no token flow to check.not applicable - 18Sign-in metadata
oauth-url-schemeChecks the published sign-in metadata for unsafe URLs.No sign-in (OAuth) metadata is published, so there are no sign-in addresses to check.not applicable
Source: our own test, 2026-10-23, mcp-runtime-engine@0.1.0. Which checks count and why follows our published scoring model.
What happened during the test?
Feature #8
The steps in order, as the test record holds them.
- 0 msStarted a read-only session with the server's remote endpoint
- 1,028 msConnected after 0 redirects. No credential was sent or needed.
- Read the inventory: 31 tools, 4 resources, 0 prompts.
- 5,072 msFinished: 3 checks exercised.
Reference
How is it classified?
18 fields
One value per field, each with its source. Fields written by AI models are marked. A field with no value says why.
“paid per call”
“Not stated” means the source did not state it; “does not apply” means it does not apply to this kind of server; “our limit” means our own lookup failed. A linked value opens the directory filtered to the same value.
Cite or correct this page
Sources and corrections
Suggested citation
ProtocolProbe. “dev.shveik/agdata: MCP server record.” Data checked 2026-10-23. https://protocolprobe.com/mcp/servers/dev.shveik/agdata
Figures are valid for the date shown. Our facts on this page are CC BY 4.0: reuse them with a link and the check date. Quoted text (the publisher’s descriptions, website quotes, tool names) is not ours to license.
As JSON: this record in the API · API documentation.
Badge for your README
[](https://protocolprobe.com/mcp/servers/dev.shveik/agdata)
It shows this server's record-check count: facts from public records, no tool called. It is not a security rating or a pass. It links here and updates after each data reload.
Something wrong, or is this yours?
If a fact here is wrong, tell us and we will re-check it. If you operate this server, you can run our checks on it yourself with the free self-test command; we call no server’s tools ourselves for anything shown here.
Write to hello@protocolprobe.com · corrections log · methodology