io.github.agentwares/mcp-liveness is an MCP server published by io.github.agentwares, hosted at agentwares-mcp-liveness.vercel.app. When last checked on 2026-11-21 it was working: it answers and lists its tools. It declares 5 tools, including mcp_liveness_check, mcp_liveness_explain_outcomes, mcp_liveness_drift, mcp_liveness_check_auth, mcp_liveness_check_apps.
Does an MCP server work for a stock client? By registry name or URL. Tool drift too. Free, no key.
105 connections through 5 shared values. Tap any node.
Inferred from the host name and headers: what a server runs on, not who runs it.
Makes one handshake with an MCP server's endpoint — `server/discover` at protocol revision 2026-07-28, then `initialize` for a server on the older revision — with no credential, and reports what a stock client gets. The server is named either by its exact official-registry `name`, or by its endpoint `url` when it is not in the registry or is listed under another name; exactly one of the two is accepted. Outcomes: `usable` (it connected and negotiated a protocol version), `needs_credential` (401 or 403 — by name, the answer also says whether a credential can be obtained in-band or whether a person has to create one), `charged` (402, access is for sale), `unreachable` (not an MCP server there: wrong URL, dead host, or a 200 that is not JSON-RPC), and, for a name only, `local_only` (no remote endpoint; it must be spawned as a process) and `not_listed` (no server with that exact name; its endpoint can still be checked by `url`). A `url` must be public https: private, loopback, link-local, multicast and cloud-metadata addresses are refused, redirects are re-checked, and URL checks are limited to 30 a minute per caller. Useful before relying on a server: roughly two in five remote registry listings do not answer a stock client.
| Parameter | Type | Description |
|---|---|---|
| url | string |
None yet. Changes appear after a server's second daily check.
Squares are this server's values; dots are other servers sharing them. A shared address or publisher is a lead, not proof of a common operator.
| The server's Streamable HTTP endpoint, e.g. `https://example.com/mcp`, for a server that is not in the registry or is named differently there. Public https only. Pass this or `name`, not both. |
| name | string | The server's exact name in the official MCP registry, namespaced — e.g. `io.github.owner/server`. Not a title. Pass this or `url`, not both. |
| skipCredentialCheck | boolean | Name checks only: skip the OAuth discovery chain on a 401. Faster (one request instead of up to five) but the answer then cannot say whether a credential is obtainable. A URL check never walks the chain. |
{
"name": "mcp_liveness_check",
"title": "Can a stock client use this MCP server?",
"description": "Makes one handshake with an MCP server's endpoint — `server/discover` at protocol revision\n2026-07-28, then `initialize` for a server on the older revision — with no credential, and\nreports what a stock client gets. The server is named either by its exact official-registry\n`name`, or by its endpoint `url` when it is not in the registry or is listed under another\nname; exactly one of the two is accepted.\n\nOutcomes: `usable` (it connected and negotiated a protocol version), `needs_credential`\n(401 or 403 — by name, the answer also says whether a credential can be obtained in-band\nor whether a person has to create one), `charged` (402, access is for sale),\n`unreachable` (not an MCP server there: wrong URL, dead host, or a 200 that is not\nJSON-RPC), and, for a name only, `local_only` (no remote endpoint; it must be spawned as a\nprocess) and `not_listed` (no server with that exact name; its endpoint can still be\nchecked by `url`).\n\nA `url` must be public https: private, loopback, link-local, multicast and cloud-metadata\naddresses are refused, redirects are re-checked, and URL checks are limited to 30 a\nminute per caller. Useful before relying on a server: roughly two in five remote registry\nlistings do not answer a stock client.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"properties": {
"url": {
"type": "string",
"maxLength": 2048,
"minLength": 1,
"description": "The server's Streamable HTTP endpoint, e.g. `https://example.com/mcp`, for a server that is not in the registry or is named differently there. Public https only. Pass this or `name`, not both."
},
"name": {
"type": "string",
"minLength": 1,
"description": "The server's exact name in the official MCP registry, namespaced — e.g. `io.github.owner/server`. Not a title. Pass this or `url`, not both."
},
"skipCredentialCheck": {
"type": "boolean",
"description": "Name checks only: skip the OAuth discovery chain on a 401. Faster (one request instead of up to five) but the answer then cannot say whether a credential is obtainable. A URL check never walks the chain."
}
},
"additionalProperties": false
},
"annotations": {
"title": "Check an MCP server's liveness",
"readOnlyHint": true,
"openWorldHint": true,
"idempotentHint": false,
"destructiveHint": false
}
}Returns the six outcomes this server can report — usable, needs_credential, charged, unreachable, local_only and not_listed — what each means for an agent choosing a server from the registry, and the usual next move for each. A static reference: it makes no network request and its answer never changes between calls.
No parameters.
{
"name": "mcp_liveness_explain_outcomes",
"title": "What the outcomes mean",
"description": "Returns the six outcomes this server can report — usable, needs_credential, charged, unreachable, local_only and not_listed — what each means for an agent choosing a server from the registry, and the usual next move for each. A static reference: it makes no network request and its answer never changes between calls.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"properties": {},
"additionalProperties": false
},
"annotations": {
"title": "Explain liveness outcomes",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Given a server's exact name in the official MCP registry, reports which of its tools were added, removed, or had their description, input schema or output schema change between a named month and the latest monthly snapshot. Useful when a server that worked before behaves differently now, when a prompt or an integration written against its tools has started failing, or before relying on a cached tool description. A liveness check says a server answers today; this says whether it is still the same server that code was written against. The answer names the changed tools and the kind of change. It does not diff the schemas themselves: the snapshots keep content hashes, not bodies, so `detail` hands back the release-asset location of both full months for a caller that needs the exact diff. A month with no snapshot for that server gets an error naming the months that do have one, never an empty diff that would read as `nothing changed`.
| Parameter | Type | Description |
|---|---|---|
| name * | string | Exact registry name, namespaced — e.g. `io.github.owner/server`. Not a URL and not a title. |
| since | string | The month to diff from, as `YYYY-MM`. Omit it to diff from the earliest month this server appears in, which is the stable choice — a hard-coded month goes wrong as soon as the next snapshot lands. |
{
"name": "mcp_liveness_drift",
"title": "What changed in this server's tools since a given month",
"description": "Given a server's exact name in the official MCP registry, reports which of its tools were\nadded, removed, or had their description, input schema or output schema change between\na named month and the latest monthly snapshot.\n\nUseful when a server that worked before behaves differently now, when a prompt or an\nintegration written against its tools has started failing, or before relying on a cached\ntool description. A liveness check says a server answers today; this says whether it is\nstill the same server that code was written against.\n\nThe answer names the changed tools and the kind of change. It does not diff the schemas\nthemselves: the snapshots keep content hashes, not bodies, so `detail` hands back the\nrelease-asset location of both full months for a caller that needs the exact diff.\n\nA month with no snapshot for that server gets an error naming the months that do have\none, never an empty diff that would read as `nothing changed`.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"name"
],
"properties": {
"name": {
"type": "string",
"minLength": 1,
"description": "Exact registry name, namespaced — e.g. `io.github.owner/server`. Not a URL and not a title."
},
"since": {
"type": "string",
"pattern": "^\\d{4}-\\d{2}$",
"description": "The month to diff from, as `YYYY-MM`. Omit it to diff from the earliest month this server appears in, which is the stable choice — a hard-coded month goes wrong as soon as the next snapshot lands."
}
},
"additionalProperties": false
},
"annotations": {
"title": "Diff an MCP server's tools by month",
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}Walks a remote MCP server's OAuth sign-in chain from outside, with no credential and no client registration, and returns the first step that breaks, whose it is (`server`, `identity_provider`, or `client` for a known client bug, with its issue URL), the fix, and the spec section that requires it. Steps, in order: one unauthenticated request (it must answer 401 with a WWW-Authenticate challenge); RFC 9728 protected-resource metadata; RFC 8414 or OpenID Connect discovery at the locations the 2026-07-28 MCP spec lists, with the issuer and PKCE S256 checked; and how a client gets a client ID (client ID metadata documents, dynamic registration, or a pasted ID). Then a verdict per client — claude.ai, Claude Code, ChatGPT, VS Code, Cursor — emulated from each client's published requirements and known issues; no client is run. Use it when a user says a connector shows an error such as 'Authorization with the MCP server failed' or 'Dynamic Client Registration not supported', or before publishing a server that needs sign-in. `verdict` is ok, at_risk, broken, no_auth (it asked for no credential) or unreachable. The URL must be public https; checks are limited to 30 a minute per caller.
| Parameter | Type | Description |
|---|---|---|
| url * | string | The server's Streamable HTTP endpoint, as a user would enter it in a client, e.g. `https://example.com/mcp`. Public https only. |
| client | string | Judge for one client only: `claude-ai` (claude.ai and Claude Desktop), `claude-code`, `chatgpt`, `vscode` or `cursor`. Omit to get all five. |
{
"name": "mcp_liveness_check_auth",
"title": "Where does this MCP server's sign-in break?",
"description": "Walks a remote MCP server's OAuth sign-in chain from outside, with no credential and no\nclient registration, and returns the first step that breaks, whose it is (`server`,\n`identity_provider`, or `client` for a known client bug, with its issue URL), the fix, and\nthe spec section that requires it.\n\nSteps, in order: one unauthenticated request (it must answer 401 with a WWW-Authenticate\nchallenge); RFC 9728 protected-resource metadata; RFC 8414 or OpenID Connect discovery at\nthe locations the 2026-07-28 MCP spec lists, with the issuer and PKCE S256 checked; and how a\nclient gets a client ID (client ID metadata documents, dynamic registration, or a pasted ID).\nThen a verdict per client — claude.ai, Claude Code, ChatGPT, VS Code, Cursor — emulated from\neach client's published requirements and known issues; no client is run.\n\nUse it when a user says a connector shows an error such as 'Authorization with the MCP\nserver failed' or 'Dynamic Client Registration not supported', or before publishing a server\nthat needs sign-in. `verdict` is ok, at_risk, broken, no_auth (it asked for no credential) or\nunreachable. The URL must be public https; checks are limited to 30 a minute per caller.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"url"
],
"properties": {
"url": {
"type": "string",
"maxLength": 2048,
"minLength": 1,
"description": "The server's Streamable HTTP endpoint, as a user would enter it in a client, e.g. `https://example.com/mcp`. Public https only."
},
"client": {
"enum": [
"claude-ai",
"claude-code",
"chatgpt",
"vscode",
"cursor"
],
"type": "string",
"description": "Judge for one client only: `claude-ai` (claude.ai and Claude Desktop), `claude-code`, `chatgpt`, `vscode` or `cursor`. Omit to get all five."
}
},
"additionalProperties": false
},
"annotations": {
"title": "Check an MCP server's OAuth sign-in",
"readOnlyHint": true,
"openWorldHint": true,
"idempotentHint": false,
"destructiveHint": false
}
}Checks a remote MCP server's MCP Apps UI (SEP-1865 `ui://` resources) from outside, with no credential and without calling any tool, and returns the first step that stops the view rendering, whose it is (`server`, or `host` for a known host bug with its issue URL), the fix, and the source (the spec, or the host's published page, with the date read). It opens the server as an MCP Apps host would, then reads tools/list, resources/list and resources/read: each tool's link to its view (_meta.ui.resourceUri, or ChatGPT's openai/outputTemplate alias), the view's MIME type (text/html;profile=mcp-app) and HTML, what the HTML loads against the CSP a host builds from _meta.ui.csp, whether it sends ui/initialize, ui.domain for Claude and for ChatGPT, and what OpenAI's directory review checks (explicit tool annotations, descriptions, a public https endpoint, the domain-verification file). Verdicts per host — `spec`, `claude`, `chatgpt`, `chatgpt-directory` — are judged from each host's published pages; no host is run. Use it when a user says an MCP App or ChatGPT app shows a blank or invisible frame, 'Resource not found: ui://', 'Invalid ui.domain format', or a submission error such as annotations_required. `verdict` is ok, at_risk, broken, no_ui, needs_sign_in or unreachable. The URL must be public https; checks are limited to 30 a minute per caller.
| Parameter | Type | Description |
|---|---|---|
| url * | string | The server's Streamable HTTP endpoint, exactly as a user adds it to a host, e.g. `https://example.com/mcp` (Claude's ui.domain is a hash of this exact string). Public https only. |
| host | string | Judge for one host only: `spec` (any MCP Apps host), `claude` (claude.ai and Claude Desktop), `chatgpt`, or `chatgpt-directory` (OpenAI's review of a submitted app). Omit to get all four. |
{
"name": "mcp_liveness_check_apps",
"title": "Why does this MCP App render blank?",
"description": "Checks a remote MCP server's MCP Apps UI (SEP-1865 `ui://` resources) from outside, with no\ncredential and without calling any tool, and returns the first step that stops the view\nrendering, whose it is (`server`, or `host` for a known host bug with its issue URL), the\nfix, and the source (the spec, or the host's published page, with the date read).\n\nIt opens the server as an MCP Apps host would, then reads tools/list, resources/list and\nresources/read: each tool's link to its view (_meta.ui.resourceUri, or ChatGPT's\nopenai/outputTemplate alias), the view's MIME type (text/html;profile=mcp-app) and HTML,\nwhat the HTML loads against the CSP a host builds from _meta.ui.csp, whether it sends\nui/initialize, ui.domain for Claude and for ChatGPT, and what OpenAI's directory review\nchecks (explicit tool annotations, descriptions, a public https endpoint, the\ndomain-verification file). Verdicts per host — `spec`, `claude`, `chatgpt`,\n`chatgpt-directory` — are judged from each host's published pages; no host is run.\n\nUse it when a user says an MCP App or ChatGPT app shows a blank or invisible frame, 'Resource\nnot found: ui://', 'Invalid ui.domain format', or a submission error such as\nannotations_required. `verdict` is ok, at_risk, broken, no_ui, needs_sign_in or unreachable.\nThe URL must be public https; checks are limited to 30 a minute per caller.",
"inputSchema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"url"
],
"properties": {
"url": {
"type": "string",
"maxLength": 2048,
"minLength": 1,
"description": "The server's Streamable HTTP endpoint, exactly as a user adds it to a host, e.g. `https://example.com/mcp` (Claude's ui.domain is a hash of this exact string). Public https only."
},
"host": {
"enum": [
"spec",
"claude",
"chatgpt",
"chatgpt-directory"
],
"type": "string",
"description": "Judge for one host only: `spec` (any MCP Apps host), `claude` (claude.ai and Claude Desktop), `chatgpt`, or `chatgpt-directory` (OpenAI's review of a submitted app). Omit to get all four."
}
},
"additionalProperties": false
},
"annotations": {
"title": "Check an MCP server's MCP Apps UI",
"readOnlyHint": true,
"openWorldHint": true,
"idempotentHint": false,
"destructiveHint": false
}
}Ask this server whether an MCP server can actually be used — named by its official registry name, or by its endpoint URL when it is not in the registry or is named differently there. A registry listing is a claim the publisher typed, not a connection anybody verified. This makes one handshake with the endpoint — `server/discover` at 2026-07-28, then `initialize` for an older server — with no credential, and says what a stock client gets: `usable`, `needs_credential`, `charged`, `unreachable`, and, for a name only, `local_only` or `not_listed`. A URL must be public https; private, loopback, link-local and metadata addresses are refused, and URL checks are limited to 30 a minute per caller. On a `needs_credential` answer it follows the discovery chain the spec defines — RFC 9728 protected-resource metadata, RFC 8414 authorization-server metadata, RFC 7591 dynamic client registration — and reports whether a client nobody has heard of could register itself, or whether a person has to go and create one. Those two are both HTTP 401 and the registry lists them identically. A second question this server answers is whether a listing is still what it was. Monthly snapshots of every reachable server's tool catalogue start in September 2026, and `mcp_liveness_drift` diffs one server between a month you name and the latest — which tools were added, removed, or changed their description or schemas. Nobody can backfill a month they did not record, which is why an answer here can say `unchanged` and mean it. Every answer carries a `next` field, `{ what, url, why }` or null: when a product we run today would tell you something this one call cannot — agentcheck, which monitors a server you publish and warns you when it stops answering — `next` names it; otherwise it is null. Every liveness answer also ends with `monitor`, one line on watching the endpoint over time and the URL where a person can ask, with no account, to be emailed when the server changes. When Wellknown (wellknown.network), which checks MCP endpoints every few hours, keeps a record of the server, the answer also carries `wellknown`: `{ url, note }`, the record's page. It is matched on the endpoint, and left out when there is no record. When a connector will not sign in, `mcp_liveness_check_auth` walks the server's OAuth chain from outside — the 401 challenge, protected-resource metadata, authorization-server discovery and client registration — and returns the first broken step, whose it is (the server, its identity provider, or a known client bug with the issue URL), the fix and the spec section, with a verdict per client (claude.ai, Claude Code, ChatGPT, VS Code, Cursor) emulated from each client's published requirements. When an MCP App renders blank, `mcp_liveness_check_apps` opens the server as an MCP Apps host and reads its tools and `ui://` views without calling a tool, and returns the first step that stops the view rendering — the tool link, the MIME type, what the HTML loads against its CSP, the ui/initialize bridge, ui.domain — whose it is and the fix, with a verdict per host (any MCP Apps host, Claude, ChatGPT) and for ChatGPT's directory review. It never sends a credential, never registers a client and never retries a refusal.
Names, descriptions and schemas were written by the server and are shown as data.
Shared addresses on big platforms (Cloudflare, Vercel and the like) put unrelated servers together; those are marked shared_infrastructure. A match is a pivot, not a link between operators.