2026-09-11 · back to archive
# Necropsy 2026-09-11 [run-reprobe-20260911-110158-00011a0514]
## Verdict summary [run-reprobe-20260911-110158-00011a0514]
Of the targets with publishable-window measurements: **69** alive, **101** ailing, **13** dead. **40** servers are classed auth-required (unscored — install/startup observations only, per D3). [run-reprobe-20260911-110158-00011a0514]
**53** further targets are published as insufficient-coverage (unscored): their probes finished below the duration floor at which we trust a run and reproduced that result, so they carry no life-state either. [run-reprobe-20260911-110158-00011a0514]
**What this covers.** Campaign `campaign-20260808-040920` is complete: 275 of 276 probeable targets returned a measurement, with 1 harness failures that are reported as unmeasured and never counted as dead. The candidate pool holds 614 rows, of which 338 are registry or repository references rather than packages we can install and run — out of scope, not deceased. This is a finding about 275 servers, not about the MCP ecosystem entire, and the re-probe campaign is in progress: a target absent from this set has not been measured in the current window, which is a gap in our coverage and not a verdict on the software. [run-reprobe-20260911-110158-00011a0514]
## Most of the MCP ecosystem is not software [run-reprobe-20260911-110158-00011a0514]
**338 of 614** candidates — 55% of everything the ecosystem presents as an MCP server — are registry entries or repository references rather than packages anyone can install and run. They are not published to npm or PyPI. There is no artifact to fetch, no entry point to launch, nothing to hand a client. [run-reprobe-20260911-110158-00011a0514]
This is a finding about the ecosystem, not a caveat about our pipeline. Directory listings, awesome-lists and registry pages count these rows as servers; a developer reading a count of `614 MCP servers` is reading a number of which 55% cannot be run at all. We do not score them, we do not list them as dead, and we do not count them in any figure on this site — a repository reference is out of scope, not deceased. It is also the single largest gap between what the MCP ecosystem claims to be and what it measurably is. [run-reprobe-20260911-110158-00011a0514]
The candidate pool holds 276 installable rows today. Campaign `campaign-20260808-040920` was scoped to the 276 that existed when it started, and 275 of those returned a measurement; the remainder were discovered after the campaign began and are queued for the next window. That measured population is what every other number in this edition describes. [run-reprobe-20260911-110158-00011a0514]
## What a verdict means [run-reprobe-20260911-110158-00011a0514]
Deaddit measures whether a server installs, starts, and behaves as documented. It does NOT measure whether a server is safe, trustworthy, or free of malicious intent. A life-state of alive means the software runs, not that it should be run. [run-reprobe-20260911-110158-00011a0514]
This matters concretely in this edition. At least one target in the set below describes itself as deliberately malicious software, and it is measured and published like any other, because removing measured data is worse than publishing it and the defect would be in the framing rather than the inclusion. Any target whose own name, README or description self-identifies as malicious, adversarial, exploit or red-team tooling carries the tag *self-identified adversarial tooling — liveness only, no safety assessment* next to its life-state, everywhere that life-state appears. [run-reprobe-20260911-110158-00011a0514]
## The dead [run-reprobe-20260808-040947-0003a52637]
- **12306-search-mcp** [run-reprobe-20260808-040947-0003a52637]
- **@arabold/docs-mcp-server** [run-reprobe-20260821-144749-0000213d67]
- **@salesforce/mcp** [run-reprobe-20260810-022709-5e1e0e]
- **@sap/mdk-mcp-server** [run-reprobe-20260808-042621-00547d0af7]
- **@superblocksteam/mcp-server** [run-reprobe-20260810-022942-0da1f3]
- **@theia/ai-mcp-server** [run-reprobe-20260810-020418-68eee5]
- **@wonderwhy-er/desktop-commander** [run-reprobe-20260821-145009-0012ef44ba]
- **@xyd-js/mcp-server** [run-reprobe-20260826-110545-0003b9b82c]
- **academia-mcp** [run-reprobe-20260821-145258-003818ef5b]
- **accordo-workflow-mcp** [run-reprobe-20260821-145715-0053851d68]
- **mcp-server-sqlite-npx** [run-reprobe-20260821-150704-01110f3c58]
- **real-browser-mcp-server** [run-reprobe-20260821-150725-011321b018]
- **storybook-mcp-server** [run-reprobe-20260821-150754-011561430d]
## Still breathing [run-reprobe-20260826-110504-0000a00711]
- **4pda-mcp** [run-reprobe-20260826-110504-0000a00711]
- **@0xmonaco/mcp-server** [run-reprobe-20260910-110310-000059fce7]
- **@agent-infra/mcp-server-browser** [run-reprobe-20260808-041340-0008358042]
- **@aikidosec/mcp** [run-reprobe-20260808-041412-00096a9458]
- **@alchemy/mcp-server** [run-reprobe-20260808-041426-0010803bd6]
- **@bitwarden/mcp-server** [run-reprobe-20260821-144749-0001197a5e]
- **@cap-js/mcp-server** [run-reprobe-20260808-041631-001805266b]
- **@circleci/mcp-server-circleci** [run-reprobe-20260808-041703-001939d280]
- **@coinbase/cds-mcp-server** [run-reprobe-20260808-041738-002104b421]
- **@dbx-app/mcp-server** [run-reprobe-20260821-144749-0002bcedfb]
- **@doitintl/doit-mcp-server** [run-reprobe-20260808-041847-0025cc7a85]
- **@ehrocks/fe-mcp-server** [run-reprobe-20260808-041919-00270db701]
- **@ericthered926/duckduckgo-mcp-server** [run-reprobe-20260808-041939-002896840e]
- **@eslint/mcp** [run-reprobe-20260808-042023-00309d1808]
- **@heroku/mcp-server** [run-reprobe-20260808-042100-0032f5458a]
- **@knip/mcp** [run-reprobe-20260822-120022-35ea06]
- **@mantine/mcp-server** [run-reprobe-20260808-042159-00368f5a9c]
- **@microsoft/clarity-mcp-server** [run-reprobe-20260808-042218-0038f186d3]
- **@modelcontextprotocol/server-filesystem** [run-reprobe-20260808-042238-0039ee9f0a]
- **@modelcontextprotocol/server-memory** [run-reprobe-20260808-042302-0040ff6ec2]
- **@modelcontextprotocol/server-sequential-thinking** [run-reprobe-20260808-042320-0041fa224e]
- **@motiffcom/motiff-mcp-server** [run-reprobe-20260821-144900-00078ff951]
- **@mui/mcp** [run-reprobe-20260808-042339-0042c1f5bf]
- **@negokaz/excel-mcp-server** [run-reprobe-20260808-042341-0043c43827]
- **@notionhq/notion-mcp-server** [run-reprobe-20260808-042401-0045244272]
- **@openbnb/mcp-server-airbnb** [run-reprobe-20260821-144909-000803cbbe]
- **@phantom/mcp-server** [run-reprobe-20260827-110413-00013317c7]
- **@professional-wiki/mediawiki-mcp-server** [run-reprobe-20260808-042428-004756d690]
- **@qase/mcp-server** [run-reprobe-20260808-042454-00487aa251]
- **@sap-ux/fiori-mcp-server** [run-reprobe-20260808-042608-00536b0f0a]
- **@siemens/element-mcp** [run-reprobe-20260808-042847-0057923a06]
- **@softeria/ms-365-mcp-server** [run-reprobe-20260808-042924-0060f6e606]
- **@tocharianou/mcp-server-kibana** [run-reprobe-20260808-043237-00672c6ca2]
- **@transcend-io/mcp** [run-reprobe-20260821-144923-0009125330]
- **@transcend-io/mcp-server-docs** [run-reprobe-20260808-043443-0073d38f20]
- **@transloadit/mcp-server** [run-reprobe-20260808-043545-0078d2a120]
- **@tsmztech/mcp-server-salesforce** [run-reprobe-20260808-043604-0079a2d12e]
- **@ui5/mcp-server** [run-reprobe-20260808-043620-008070709b]
- **@upstash/context7-mcp** [run-reprobe-20260808-043709-0082e582b8]
- **@variflight-ai/variflight-mcp** [run-reprobe-20260821-144949-00115400f4]
- **a2a-exchange-mcp** [run-reprobe-20260808-044013-00912258a2]
- **a2a-mcp-server** [run-reprobe-20260808-044056-00956b4458]
- **aas-mcp** [run-reprobe-20260808-044513-01106d138b]
- **abaqus-mcp** [run-reprobe-20260902-110324-0002f24db9]
- **abs-mcp** [run-reprobe-20260821-145244-0035afb47c]
- **academic-research-mcp** [run-reprobe-20260821-145427-0042e41ef7]
- **acb-tax-mcp** [run-reprobe-20260821-145608-004743b48c]
- **acmt001-mcp** [run-reprobe-20260821-145924-006356ef05]
- **acp-mcp-server** [run-reprobe-20260821-145936-0065fa6f8d]
- **acquisition-gov-mcp** [run-reprobe-20260823-023341-0002697d00]
- **activitywatch-mcp-server-py** [run-reprobe-20260821-150154-0076d6b54a]
- **chrome-devtools-mcp** [run-reprobe-20260808-045043-012330e918]
- **codex-mcp-server** [run-reprobe-20260821-150621-0106262f41]
- **dataforseo-mcp-server** [run-reprobe-20260821-150630-01079dc15e]
- **gezhe-mcp-server** [run-reprobe-20260808-045145-012657b3fe]
- **hostinger-api-mcp** [run-reprobe-20260808-045204-01274c1a56]
- **langsmith-mcp-server** [run-reprobe-20260808-045214-0129ab16ed]
- **lsp-mcp-server** [run-reprobe-20260821-150642-0109b94dbd]
- **malicious-mcp-server** — *self-identified adversarial tooling — liveness only, no safety assessment* [run-reprobe-20260808-045228-0130805585]
- **mcp-hello-world** [run-reprobe-20260808-045236-01318a8a23]
- **mcp-server-kubernetes** [run-reprobe-20260808-045301-01334fbe43]
- **next-devtools-mcp** [run-reprobe-20260808-045312-013454c9b5]
- **playwright-stealth-mcp-server** [run-reprobe-20260908-110334-0002af7baf]
- **scryfall-mcp-server** [run-reprobe-20260907-110144-0001da74eb]
- **shadcn-ui-mcp-server** [run-reprobe-20260821-150738-0114a6de24]
- **square-mcp-server** [run-reprobe-20260808-045412-0138c49119]
- **tavily-mcp** [run-reprobe-20260808-045421-013969ec26]
- **terraform-mcp-server** [run-reprobe-20260808-045437-01406205f2]
- **touchdesigner-mcp-server** [run-reprobe-20260808-045444-0141c386da]
## The waiting room [run-reprobe-20260808-040920-0001996b90]
- **0xchron-polymarket-mcp** [run-reprobe-20260808-040920-0001996b90]
- **0xdegenmo-lighter-mcp** [run-reprobe-20260808-040943-00023a01e0]
- **1up-mcp** [run-reprobe-20260808-041125-0000eed7f6]
- **1xn-vmcp** [run-reprobe-20260808-041148-0002901d25]
- **360solutions-bc-mcp** [run-reprobe-20260808-041154-0003ceb52f]
- **3dcitydb-mcp-server** [run-reprobe-20260808-041226-000408caf5]
- **3dsmax-mcp** [run-reprobe-20260808-041254-000538d229]
- **@benborla29/mcp-server-mysql** [run-reprobe-20260810-021432-917da0]
- **@browserstack/mcp-server** [run-reprobe-20260810-021515-887120]
- **@cloudflare/mcp-server-cloudflare** [run-reprobe-20260810-022925-769297]
- **@iobroker/mcp-server** [run-reprobe-20260821-144758-00030c4827]
- **@ivotoby/openapi-mcp-server** [run-reprobe-20260810-020654-662384]
- **@kintone/mcp-server** [run-reprobe-20260821-144815-00046a5310]
- **@mapbox/mcp-server** [run-reprobe-20260810-021812-1f716e]
- **@mastra/mcp-docs-server** [run-reprobe-20260908-110303-000137280b]
- **@modelcontextprotocol/server-everything** [run-reprobe-20260821-144842-0006fd7159]
- **@nexus2520/jira-mcp-server** [run-reprobe-20260910-110343-0002e516b1]
- **@onozaty/redmine-mcp-server** [run-reprobe-20260909-110221-00015eb7d7]
- **@pandacss/mcp** [run-reprobe-20260808-042418-00466f64a8]
- **@supabase/mcp-server-postgrest** [run-reprobe-20260810-022435-292d93]
- **@taazkareem/clickup-mcp-server** [run-reprobe-20260810-020841-0fc05e]
- **@transcend-io/mcp-server-base** [run-reprobe-20260810-023128-eebc94]
- **@unthread-io/mcp-server** [run-reprobe-20260828-110139-000135df5c]
- **@xeroapi/xero-mcp-server** [run-reprobe-20260808-043843-00862d9b0e]
- **@zapier/zapier-sdk-mcp** [run-reprobe-20260902-110303-000170cb4a]
- **a11y-mcp** [run-reprobe-20260808-043916-0089ce5d90]
- **a2a-mcp** [run-reprobe-20260808-044037-0093ae2fcb]
- **a2a-raid-mcp** [run-reprobe-20260808-044147-0097b385b8]
- **a2c-smcp** [run-reprobe-20260808-044227-0101c5592f]
- **a360-mcp** [run-reprobe-20260821-145032-0025a0f7f3]
- **a3m-mcp-server** [run-reprobe-20260821-145035-0026fa7d40]
- **aa-mcp-server** [run-reprobe-20260808-044323-010416a599]
- **aaaa-nexus-mcp** [run-reprobe-20260808-044329-01051ec64b]
- **aaep-mcp-bridge** [run-reprobe-20260808-044404-0106828557]
- **aagentic-terminal-mcp-prasad** [run-reprobe-20260808-044407-010707137d]
- **aapanel-mcp** [run-reprobe-20260808-044427-0108c19896]
- **aas-mcp-server** [run-reprobe-20260814-110115-21b845]
- **abbe-mcp** [run-reprobe-20260808-044627-01136dd230]
- **abi-to-mcp** [run-reprobe-20260808-044700-0114ec6191]
- **able-mcp-core** [run-reprobe-20260808-044744-0115d0975c]
- **able-mcp-notable** [run-reprobe-20260808-044811-0116d3b8cc]
- **ableton-live-mcp** [run-reprobe-20260808-044912-012034e213]
- **ableton-live-mcp-server** [run-reprobe-20260821-145058-0027e81bc9]
- **ableton-mcp-lofifren** [run-reprobe-20260821-145102-0028086d6f]
- **ableton-mcp-server** [run-reprobe-20260821-145135-002965ed81]
- **ableton-mcp-suite** [run-reprobe-20260821-145141-00302b3589]
- **ableton-osc-mcp** [run-reprobe-20260821-145144-00318e19ed]
- **abraflexi-mcp-server** [run-reprobe-20260821-145207-003291176e]
- **abrasio-mcp** [run-reprobe-20260821-145213-00332aa201]
- **abroad-mcp-server-qiangyang** [run-reprobe-20260821-145221-00345dc349]
- **ac-jobs-mcp** [run-reprobe-20260910-110400-00032c4194]
- **ac-mcp-server** [run-reprobe-20260821-145251-0037a099ab]
- **academic-assistant-mcp** [run-reprobe-20260821-145348-0039d6f5ba]
- **academic-figures-mcp** [run-reprobe-20260821-145406-00403777b1]
- **academic-mcp** [run-reprobe-20260821-145416-0041f1de10]
- **academic-search-mcp** [run-reprobe-20260821-145449-0043ba9454]
- **academy-mcp** [run-reprobe-20260821-145539-0044669571]
- **acatome-mcp** [run-reprobe-20260821-145548-00454ed27f]
- **acatome-quest-mcp** [run-reprobe-20260821-145553-0046c2bd92]
- **acc-c2c-mcp** [run-reprobe-20260821-145638-00484b8868]
- **accela-mcp** [run-reprobe-20260821-145642-004904e0e1]
- **accessibility-ai-mcp** [run-reprobe-20260821-145647-0050ef32ed]
- **accessibility-mcp-server** [run-reprobe-20260821-145709-0051a086cd]
- **accommodation-expenses-mcp** [run-reprobe-20260821-145712-00528e251f]
- **accounting-ai-mcp** [run-reprobe-20260821-145739-00548f4b8f]
- **accounting-subject-analysis-mcp-server** [run-reprobe-20260821-145747-00557a190d]
- **aceflow-mcp-server** [run-reprobe-20260821-145804-0057472008]
- **acemcp** [run-reprobe-20260821-145817-00584ba2d4]
- **acemcp-fast** [run-reprobe-20260821-145840-005916115b]
- **aci-mcp** [run-reprobe-20260821-145851-0061ab90c3]
- **acli-mcp-server** [run-reprobe-20260821-145900-006253f057]
- **acp-mcp** [run-reprobe-20260821-145925-0064272142]
- **actionai-mcp** [run-reprobe-20260821-150027-00681a2e1f]
- **actionlayer-mcp** [run-reprobe-20260821-150038-0069a296c5]
- **actions-latest-mcp** [run-reprobe-20260821-150041-0070ce5a0d]
- **actionstep-mcp** [run-reprobe-20260821-150116-0072d8dec2]
- **acttrace-mcp** [run-reprobe-20260821-150211-007743cd7f]
- **actvalue-lambda-mcp-server** [run-reprobe-20260821-150216-00785931c6]
- **acumatica-mcp** [run-reprobe-20260821-150232-0079276aaf]
- **acuris-mcp** [run-reprobe-20260821-150244-0080a67c31]
- **adafruit-mcp3008** [run-reprobe-20260821-150319-0089cd8dad]
- **adafruit-mcp4725** [run-reprobe-20260821-150320-0090d6ae18]
- **adafruit-mcp9808** [run-reprobe-20260821-150321-0091b8dafc]
- **adalm2000-mcp** [run-reprobe-20260821-150344-009286589c]
- **adam-calculator-mcp** [run-reprobe-20260821-150346-0093a76dcc]
- **adam-mcp** [run-reprobe-20260821-150347-0094b595cf]
- **adam-mcp-cli** [run-reprobe-20260821-150419-0095420f08]
- **adam-mcp-py** [run-reprobe-20260821-150425-0096ad725a]
- **adaptive-agent-mcp** [run-reprobe-20260821-150451-00970d1c2c]
- **adb-mcp** [run-reprobe-20260821-150524-01001a08e9]
- **adb-mcp-server** [run-reprobe-20260821-150533-01014f2c3d]
- **adbpg-mcp-server** [run-reprobe-20260821-150550-01033e17df]
- **adbpg-mcp-server-test** [run-reprobe-20260821-150600-010431358a]
- **adcirc-mcp** [run-reprobe-20260821-150604-0105fc994d]
- **agentation-mcp** [run-reprobe-20260810-020729-03393f]
- **argocd-mcp** [run-reprobe-20260810-021637-f7717d]
- **docusaurus-plugin-mcp-server** [run-reprobe-20260808-045053-0125f8faab]
- **mcp-searxng** [run-reprobe-20260810-020950-2b2a62]
- **openapi-mcp-generator** [run-reprobe-20260821-150716-01122e1d75]
- **polaris-mcp-server** [run-reprobe-20260810-020718-ea295f]
- **prometheus-mcp** [run-reprobe-20260903-110304-000257fd8a]
## Auth-required (unscored) [run-reprobe-20260810-022458-44ce44]
Refused a key honestly rather than failing — measured, not broken. These servers carry no life-state: a server that declines to start without credentials has told us about its auth model, not about its health. [run-reprobe-20260810-022458-44ce44]
- **@amap/amap-maps-mcp-server** [run-reprobe-20260810-022458-44ce44]
- **@apify/actors-mcp-server** [run-reprobe-20260810-022604-02f0e4]
- **@azure-devops/mcp** [run-reprobe-20260810-021944-d345cf]
- **@brave/brave-search-mcp-server** [run-reprobe-20260810-022244-cd8d9f]
- **@contentful/mcp-server** [run-reprobe-20260808-041743-0022a63862]
- **@currents/mcp** [run-reprobe-20260810-021219-513633]
- **@delorenj/mcp-server-trello** [run-reprobe-20260810-022349-86ea96]
- **@dynatrace-wwse/dynatrace-mcp-server-otel** [run-reprobe-20260808-041910-00264a3440]
- **@esaio/esa-mcp-server** [run-reprobe-20260810-020812-53fd36]
- **@gongrzhe/server-gmail-autoauth-mcp** [run-reprobe-20260810-021016-f874b1]
- **@hubspot/mcp-server** [run-reprobe-20260810-022903-c85dee]
- **@nexus2520/bitbucket-mcp-server** [run-reprobe-20260808-042352-0044ce2b55]
- **@ragieai/mcp-server** [run-reprobe-20260808-042506-00499ef4e1]
- **@roychri/mcp-server-asana** [run-reprobe-20260810-020614-ad5765]
- **@runpod/mcp-server** [run-reprobe-20260808-042530-00513a34b6]
- **@sentry/mcp-server** [run-reprobe-20260808-042745-0055127261]
- **@shortcut/mcp** [run-reprobe-20260810-022535-118396]
- **@sigmacomputing/slack-mcp-server** [run-reprobe-20260810-022512-4af4ed]
- **@supabase/mcp-server-supabase** [run-reprobe-20260808-043030-00631bc4cc]
- **@transcend-io/mcp-server-admin** [run-reprobe-20260810-021706-40dda0]
- **@transcend-io/mcp-server-assessment** [run-reprobe-20260810-022138-eb5d80]
- **@transcend-io/mcp-server-consent** [run-reprobe-20260808-043405-00710990ac]
- **@transcend-io/mcp-server-discovery** [run-reprobe-20260810-022211-341aa5]
- **@transcend-io/mcp-server-dsr** [run-reprobe-20260808-043450-00749eadf8]
- **@transcend-io/mcp-server-inventory** [run-reprobe-20260808-043509-00752e095f]
- **@transcend-io/mcp-server-preferences** [run-reprobe-20260810-021743-84d535]
- **@transcend-io/mcp-server-workflows** [run-reprobe-20260810-022321-8ca908]
- **@upstash/mcp-server** [run-reprobe-20260808-043712-0083b9a18b]
- **@z_ai/mcp-server** [run-reprobe-20260810-020752-d4e5bd]
- **@zencoderai/slack-mcp-server** [run-reprobe-20260911-110158-00011a0514]
- **@zereight/mcp-gitlab** [run-reprobe-20260903-110233-00018c8348]
- **a2a4b2b-mcp** [run-reprobe-20260808-044154-00989d42bd]
- **abacum-mcp-server** [run-reprobe-20260808-044612-0112a3a6c6]
- **active-claude-github-mcp** [run-reprobe-20260821-150116-0073690568]
- **activecollab-mcp** [run-reprobe-20260821-150145-00759fad73]
- **adaptive-mcp** [run-reprobe-20260821-150453-00988bb7f0]
- **deepl-mcp-server** [run-reprobe-20260821-150639-0108adb420]
- **kubernetes-mcp-server** [run-reprobe-20260810-021615-f42b0d]
- **serper-search-scrape-mcp-server** [run-reprobe-20260810-021626-01ae83]
- **slite-mcp-server** [run-reprobe-20260810-021244-02c5fc]
## Insufficient coverage (unscored) [run-reprobe-20260811-011629-f888d5]
These targets failed faster than the 8-second floor below which we do not trust that the harness genuinely ran. A fast failure and a harness that never started look identical from the outside, and mistaking one for the other is exactly what produced this project's false-death incident. Each was re-probed with verbose logging and reproduced its failure cause; that reproduction is what makes it a property of the target rather than of the instrument. They carry no life-state — a run we do not trust yields no verdict. [run-reprobe-20260811-011629-f888d5]
- **021-mcp** — installs but exposes no runnable entry point (29.0s) [run-reprobe-20260811-011629-f888d5]
- **1wlf-tms-mcp** — installs but exposes no runnable entry point (23.1s) [run-reprobe-20260811-011529-d1641f]
- **3tears-mcp** — install failed — npm error code E404 npm error 404 Not Found - GET https://registry.npmjs.org/3tears-mcp - Not found npm error 404 npm error 404 '3tears-mcp@*' is not in this re (2.0s) [run-reprobe-20260810-022531-6c5c51]
- **5g-ddos-mcp** — install failed — WARNING: typer 0.27.1 does not provide the extra 'all' (44.1s) [run-reprobe-20260811-011840-8a04ca]
- **@azure/mcp** — unmeasured — probe aborted or did not respond (duration unrecorded) [run-reprobe-20260810-022411-78f5f8]
- **@smartbear/mcp** — unmeasured — probe aborted or did not respond (duration unrecorded) [run-reprobe-20260810-022026-2295f1]
- **@storybook/mcp** — installs but exposes no runnable entry point (7.0s) [run-reprobe-20260810-021423-cfc2f8]
- **@winor30/mcp-server-datadog** — unmeasured — probe aborted or did not respond (duration unrecorded) [run-reprobe-20260810-021104-bd7c58]
- **a11y-expert-mcp** — installs but exposes no runnable entry point (33.1s) [run-reprobe-20260811-011554-dad313]
- **a207-clinical-calc-mcp** — install failed — no matching distribution for the cage's Python (2.2s) [run-reprobe-20260821-145013-0013444fcc]
- **a207-followup-mcp** — install failed — no matching distribution for the cage's Python (2.2s) [run-reprobe-20260821-145015-00141a3cd4]
- **a207-gamification-mcp** — install failed — no matching distribution for the cage's Python (2.2s) [run-reprobe-20260821-145017-00159b5bde]
- **a207-his-mcp** — install failed — no matching distribution for the cage's Python (2.0s) [run-reprobe-20260821-145018-001667a497]
- **a207-knowledge-mcp** — install failed — no matching distribution for the cage's Python (2.1s) [run-reprobe-20260821-145020-00176704f2]
- **a207-lis-mcp** — install failed — no matching distribution for the cage's Python (2.1s) [run-reprobe-20260821-145021-0018a77e68]
- **a207-meal-plan-mcp** — install failed — no matching distribution for the cage's Python (2.3s) [run-reprobe-20260821-145023-00198b2096]
- **a207-notification-mcp** — install failed — no matching distribution for the cage's Python (2.3s) [run-reprobe-20260821-145025-0020016dc2]
- **a207-nutrition-assessment-mcp-nfyy** — install failed — no matching distribution for the cage's Python (2.3s) [run-reprobe-20260821-145026-0021ac6a19]
- **a207-nutrition-calc-mcp** — install failed — no matching distribution for the cage's Python (2.4s) [run-reprobe-20260821-145028-0022146dc1]
- **a207-report-mcp** — install failed — no matching distribution for the cage's Python (2.2s) [run-reprobe-20260821-145029-0023d32d4b]
- **a207-risk-rules-mcp** — install failed — no matching distribution for the cage's Python (2.1s) [run-reprobe-20260821-145031-0024bfd0bb]
- **a2a-dm-mcp** — installs but exposes no runnable entry point (29.5s) [run-reprobe-20260811-015732-8a4e38]
- **a2a-governance-bridge-mcp** — installs but exposes no runnable entry point (58.6s) [run-reprobe-20260811-011428-f09049]
- **a2a-mcp-connector** — install failed — Command '['/usr/local/bin/python3', '-m', 'pip', 'install', '--quiet', '--no-input', '--target', '/probe-scratch/site-packages', 'a2a-mcp-connector']' timed out (180.1s) [run-reprobe-20260811-020033-b1ebbf]
- **a2a-protocol-mcp-server** — installs but exposes no runnable entry point (28.9s) [run-reprobe-20260811-011809-7f9a2c]
- **a2amcp-sdk** — install failed — npm error code E404 npm error 404 Not Found - GET https://registry.npmjs.org/a2amcp-sdk - Not found npm error 404 npm error 404 'a2amcp-sdk@*' is not in this re (2.1s) [run-reprobe-20260810-021739-0309f5]
- **a2anet-mcp** — install failed — npm error code E404 npm error 404 Not Found - GET https://registry.npmjs.org/a2anet-mcp - Not found npm error 404 npm error 404 'a2anet-mcp@*' is not in this re (2.1s) [run-reprobe-20260810-022701-c9fd9d]
- **a2cr-mcp** — unmeasured — probe aborted or did not respond (duration unrecorded) [run-reprobe-20260814-110004-9e6bed]
- **aa-mcp** — installs but exposes no runnable entry point (66.9s) [run-reprobe-20260811-011700-7539c5]
- **aapt-mcp-server** — installs but exposes no runnable entry point (27.3s) [run-reprobe-20260811-020004-3e04b9]
- **able-mcp-notable-96** — installs but failed to start (29.2s) [run-reprobe-20260811-015933-1decb2]
- **able-mcp-suitable** — installs but failed to start (27.0s) [run-reprobe-20260811-011926-828a50]
- **ableton-auto-mix-mcp** — install failed — no diagnostic (86.7s) [run-reprobe-20260811-015803-5cd083]
- **abuseipdb-mcp** — installs but exposes no runnable entry point (2.1s) [run-reprobe-20260821-145248-00366b7fbb]
- **accton-mcpdemo-weather** — install failed — no matching distribution for the cage's Python (1.7s) [run-reprobe-20260821-145801-005633c215]
- **acex-mcp-server** — install failed — no matching distribution for the cage's Python (1.8s) [run-reprobe-20260821-145848-0060281bbc]
- **acrf-mcp-scan** — installs but failed to start (2.5s) [run-reprobe-20260821-150015-0066c2273d]
- **acsearch-mcp** — unmeasured — probe aborted or did not respond (duration unrecorded) [run-reprobe-20260821-150018-0067d750a6]
- **actionsmcp** — installs but exposes no runnable entry point (3.1s) [run-reprobe-20260821-150112-007136419e]
- **activecamp-mcp** — unmeasured — probe aborted or did not respond (duration unrecorded) [run-reprobe-20260821-150131-0074abfb22]
- **adafruit-circuitpython-mcp230xx** — installs but exposes no runnable entry point (6.9s) [run-reprobe-20260821-150249-0081bc7499]
- **adafruit-circuitpython-mcp2515** — installs but exposes no runnable entry point (6.8s) [run-reprobe-20260821-150256-0082fb6aa8]
- **adafruit-circuitpython-mcp3421** — installs but exposes no runnable entry point (6.9s) [run-reprobe-20260821-150257-0083c77cf4]
- **adafruit-circuitpython-mcp3xxx** — installs but exposes no runnable entry point (7.0s) [run-reprobe-20260821-150304-0084a29c73]
- **adafruit-circuitpython-mcp4725** — installs but exposes no runnable entry point (7.0s) [run-reprobe-20260821-150305-00850cba85]
- **adafruit-circuitpython-mcp4728** — installs but exposes no runnable entry point (6.9s) [run-reprobe-20260821-150312-0086bd737b]
- **adafruit-circuitpython-mcp9600** — installs but exposes no runnable entry point (6.9s) [run-reprobe-20260821-150313-0087066eae]
- **adafruit-circuitpython-mcp9808** — installs but exposes no runnable entry point (7.0s) [run-reprobe-20260821-150313-0088df8f08]
- **adaptorch-mcp** — install failed — ERROR: Cannot install adaptorch-mcp==0.3.0, adaptorch-mcp==0.4.1 and adaptorch-mcp==0.4.2 because these package versions have conflicting dependencies. ERROR: R (2.3s) [run-reprobe-20260821-150521-0099078fb9]
- **adb-mysql-mcp-server** — install failed — no matching distribution for the cage's Python (1.9s) [run-reprobe-20260821-150547-0102ebea61]
- **datadog-mcp-server** — unmeasured — probe aborted or did not respond (duration unrecorded) [run-reprobe-20260810-021307-c1763d]
- **mcp-server** — unmeasured — probe aborted or did not respond (duration unrecorded) [run-reprobe-20260821-150704-0110abf1db]
- **playwright-mcp-server** — unmeasured — probe aborted or did not respond (duration unrecorded) [run-reprobe-20260822-060140-00027d2937]
## Claimed vs actual tools [run-reprobe-20260911-110158-00011a0514]
A server's README is a promise about its tool surface. We compared what each package documents against what it actually returned from `tools/list` at handshake. Of the **24** servers that both completed a handshake and document a tool surface at all, **10** document at least one tool they do not expose, and **4** expose tools they never document. [run-reprobe-20260911-110158-00011a0514]
Documented, not exposed [run-reprobe-20260808-045043-012330e918]:
- **chrome-devtools-mcp** documents `click_at`, `close_heapsnapshot`, `compare_heapsnapshots`, `execute_3p_developer_tool`, `execute_webmcp_tool`, `get_heapsnapshot_class_nodes` (+17 more), and exposes none of them [run-reprobe-20260808-045043-012330e918]
- **hostinger-api-mcp** documents `additional_details`, `datacenter_code`, `from_line`, `is_patchable`, `is_patching_in_progress`, `item_id` (+8 more), and exposes none of them [run-reprobe-20260808-045204-01274c1a56]
- **langsmith-mcp-server** documents `as_of`, `max_chars_per_page`, `page_number`, `preview_chars`, `reference_dataset_id`, `reference_dataset_name` (+3 more), and exposes none of them [run-reprobe-20260808-045214-0129ab16ed]
- **next-devtools-mcp** documents `get_errors`, `get_logs`, `get_page_metadata`, `get_project_metadata`, `get_server_action_by_id`, `project_path`, and exposes none of them [run-reprobe-20260808-045312-013454c9b5]
- **mcp-server-kubernetes** documents `cleanup_pods`, `helm_install`, `helm_template_apply`, `helm_template_uninstall`, and exposes none of them [run-reprobe-20260808-045301-01334fbe43]
- **@doitintl/doit-mcp-server** documents `layer_ids`, `scheme_ids`, `ss_id`, and exposes none of them [run-reprobe-20260808-041847-0025cc7a85]
- **@agent-infra/mcp-server-browser** documents `browser_vision_screen_capture`, `browser_vision_screen_click`, and exposes none of them [run-reprobe-20260808-041340-0008358042]
- **@transloadit/mcp-server** documents `include_builtin`, `include_content`, and exposes none of them [run-reprobe-20260808-043545-0078d2a120]
- **touchdesigner-mcp-server** documents `get_module_help`, and exposes none of them [run-reprobe-20260808-045444-0141c386da]
- **@heroku/mcp-server** documents `transfer_app`, and exposes none of them [run-reprobe-20260808-042100-0032f5458a]
Exposed, never documented [run-reprobe-20260808-042454-00487aa251]:
- **@qase/mcp-server** exposes `qase_api`, `qase_case_upsert`, `qase_ci_report`, `qase_defect_upsert`, `qase_get`, `qase_project_context` (+6 more), documented nowhere in its tool section [run-reprobe-20260808-042454-00487aa251]
- **a2cr-mcp** exposes `archive_work_thread`, `close_work_thread`, `create_work_thread`, `get_handoff`, `get_work_thread`, `list_work_threads` (+2 more), documented nowhere in its tool section [run-reprobe-20260808-044228-0102224a7c]
- **a2a-exchange-mcp** exposes `publish_handoff`, `publish_request`, `publish_response`, `read_agent`, `read_board`, documented nowhere in its tool section [run-reprobe-20260808-044013-00912258a2]
- **@modelcontextprotocol/server-filesystem** exposes `read_file`, documented nowhere in its tool section [run-reprobe-20260808-042238-0039ee9f0a]
Method, stated so the finding can be checked: a tool counts as *claimed* only when it appears as a backticked identifier inside a README section headed tool/command/capability/function. Prose mentions and config snippets are ignored. That undercounts divergence, which is the direction we prefer to be wrong in — a false accusation about someone else's software is worse than a missed one. Registry metadata was read as data, never executed. [run-reprobe-20260911-110158-00011a0514]
Coverage for this section, stated so the denominator can be audited: 46 targets completed a handshake in the publishable window; 3 published no README we could retrieve and 19 publish a README with no tool section, leaving 24 comparable. Those exclusions are reported as unmeasured, not as compliance. [run-reprobe-20260911-110158-00011a0514]
## Resurrections and relapses [run-reprobe-20260911-110158-00011a0514]
**Not yet measurable.** We are not reporting that nothing came back; we are reporting that we cannot yet tell. Resurrection detection requires two windows measured by the same validated harness, and this campaign is the first valid baseline. Runs from before the harness fix are instrument-confounded and excluded from every comparison per Strategist ruling `ruling-7e19c3b21d17` D1, so there is no admissible earlier window to compare against. The 24 targets re-measured with the validated harness on both sides sit inside this campaign, which is too short an interval to call a resurrection. From the next window forward this section reports real transitions. [run-reprobe-20260911-110158-00011a0514]
Reporting this as **None** — as an earlier draft did — would assert a measurement we did not make. A zero and an absence of measurement look identical in a table and mean opposite things. [run-reprobe-20260911-110158-00011a0514]
For completeness on the alarm raised against the first draft: it asked why no falsely-dead server flipped back to alive after the handshake fix, and it was answered. The reference server was absent from the verified window rather than failing inside it, and a fresh in-window probe returns it alive. The remaining two suspects are auth-required and therefore never carried a life-state to flip. The canonical pins are probed inside this campaign rather than in a separate run, so the question cannot recur silently. [run-reprobe-20260911-110158-00011a0514]
Comparisons against pre-fix runs (5 of 257 apparently changed) are **excluded** from the above. The harness itself changed between those measurements, so a flip there cannot be attributed to the software — it is an instrument diagnostic, and reporting it as an ecosystem finding would be exactly the error this edition was sent back to fix. [run-reprobe-20260911-110158-00011a0514]
Separately, **21** targets previously carried both a life-state and an auth-required flag — a classification the Charter forbids, since a server that refuses to start without credentials has not been measured. They now carry auth-required only. That is a correction to *our* bookkeeping, not a change in anyone's software, and it is not counted as a transition above. [run-reprobe-20260911-110158-00011a0514]
## Method notes [run-reprobe-20260911-110158-00011a0514]
- **The first verified window missed its own reference server.** Gate a pins five known-good servers as ground truth, but `@modelcontextprotocol/server-memory` was never in the candidate set at all, and `@modelcontextprotocol/server-filesystem` was absent from the first verified window — so the fleet's primary correctness gate ran without checking its own reference. A campaign that cannot see its reference server cannot distinguish a broken ecosystem from a broken instrument. The pins are probed inside this campaign, not in a separate run, and the missing pin has been added to the candidate set. [run-reprobe-20260911-110158-00011a0514]
- **43-run vs 47-score.** The earlier reprobe campaign recorded 43 runs across 41 distinct targets while the gate counted 47 non-auth scores; the difference is count-of-runs versus count-of-scored-targets, not a data-loss event. Retained here because the arithmetic still confuses on first reading. [run-reprobe-20260911-110158-00011a0514]
- **Coverage is not the candidate count.** The candidate pool holds hundreds of rows, but only those published to npm or PyPI are installable software we can run. Repository references and scraped README markup are not servers; they are counted as out of scope, never as dead. [run-reprobe-20260911-110158-00011a0514]
- Findings trace to fleet run IDs; methodology is public, rubric v1.0 is immutable. [run-reprobe-20260911-110158-00011a0514]
- Maintainers: a fix picked up by the next fleet run updates your record automatically. [run-reprobe-20260911-110158-00011a0514]Every cited claim traces to a fleet run ID. Methodology is public; rubric v1.0 is immutable.