MCP Server Discovery - October 3, 2026 (Midday Sweep)

Midday sweep over the mcp.so /servers listing (50 slugs) and mcpservers.org /all pages 1-3, both via the r.jina.ai reader proxy because direct fetches return the Cloudflare/JS shell without server data. Detail pages were fetched for every candidate; endpoints were probed for liveness. Two new business-relevant servers were catalogued, each with an integration guide.

New This Sweep

Toffu MCP - AI Marketing Agent for Ad Accounts

Hosted Streamable HTTP at mcp.toffu.ai/mcp. The vendor ships an agent-native onboarding path: POST /agent/signup returns an api_key and company_id with no email required, and a real email can be supplied to make the account human-claimable later. Eight tools are published in the machine-readable capability doc at toffu.ai/.well-known/toffu.json: send_message, get_task_result, query_campaign_performance, creative_report, the staged propose_change / apply_change / undo_change triple, and fetch_memory. The read/write split matters for operators running paid acquisition on more than one platform: an agent can propose a campaign edit, hold it for approval, and still roll it back. OAuth 2.1 + PKCE with Dynamic Client Registration is available for clients that prefer a consent flow. The endpoint returns 401 unauthenticated, confirming it is live and auth-gated.

RankOrg MCP - SEO Content Engine for Agents

Remote MCP at rankorg.com/api/mcp with OAuth 2.0 (RFC 8414 discovery, RFC 7591 Dynamic Client Registration, PKCE) or a personal access token for headless clients. Twelve tools split explicitly into read and write: getSiteContext (call first to ground everything), listContent (filter by today, scheduled, published or all), getContent, getSearchPerformance (live Google Search Console clicks, impressions, CTR, average position, top pages, top queries and a daily trend), getTopicalMap (clusters through topics and subtopics to keywords with demand, difficulty, rankability and which keywords already have scheduled or published articles), plus researchKeywords, generateArticle, updateContent, rescheduleContent, publishContent and mergeContent / unmergeContent. The vendor documents publishContent as public and effectively irreversible, to be invoked only after explicit confirmation.

Listings Verified

Directory	Status	Evidence
mcpservers.org	LISTED	Reader-proxy fetch of /servers/corpusiq-io returns HTTP 200 with title "CorpusIQ MCP Server | Awesome MCP Servers"
glama.ai/mcp	LISTED	HTTP 200 on the @corpusiq/corpusiq-docs slug
smithery.ai	REGISTRY ENTRY PRESENT	Cprusiq in the registry API response; flip-flop class, no submission path available
mcp.so	NOT LISTED	Search returns unrelated servers (Medplum, Hostinger, PLUR, Termany and others) with no CorpusIQ entry
PulseMCP	LISTED	Reader-proxy fetch of /servers/corpusiq returns the CorpusIQ listing

Identified But Not Catalogued

mcp.so /servers: Casefile (self-hosted agent case-file memory and hand-off tracker; agent-memory class already covered by the Jotter and clexo dispositions), q-ring (quantum-themed local secret keyring for coding agents; dev secrets-utility class), OrangePro (local-first behavior mapping and test generation with mutation testing; dev-qa class), Muumuu Domain MCP (first remote MCP from a Japanese domain registrar, GMO Pepabo; regional consumer service). The remaining slugs were catalogued or disposed in prior sweeps, including AON Agent Offer Network, Hostinger, BlazeCDN, Clipkit, Coinlobster, Crawlforge, Datris, EqIQ, Etincel, Farpy, Fhirhydrant, FiatDock, Foremerge, Gologin, Groundwork, Heylead, Ibanforge, Lawstronaut, Legion, LocalCan, Modelglass, Openlore, Patsnap, Perfex, SCDV, Snipara, Somacheck, Strac, Subtext, Sugra, Tracetify and OpenZiti.

mcpservers.org /all: Kinoify (media generation utility), Bond (Discord MCP; communication saturation), Buildix Hyperliquid Orderflow (crypto liquidation data), Momentra (agent identity with a trial allowance; agent infra class), Search Fragments (search-answer utility; saturated web-search class), OneFindMe (AliExpress shopping search; consumer retail class), QRSalt (QR code generator; commodity utility), OperatorNest (local stdio tooling; dev utility class). The remainder of the three pages resolved to prior-sweep dispositions, non-English marketing pages and username-fragment 404 shells.

Notes

• The mcpservers.org /all page 1 was zero-churn at roughly a two-hour gap from the morning sweep; pages 2 and 3 carried the fresh slugs. This matches the established pattern that same-day short-gap sweeps should treat /all page 1 as a cross-check and look deeper for turnover.
• The mcp.so feed's entire top block (PredictionMarketsPicks, Sooveryn, SocialAPIs, Zyte, AgentGrid.io, Builders in Fintech, FlatHunt, TATUAT.RO, Manifold, Common Paper, PixelDojo, Aayat AI) was prior dispositions from the October 1 through October 3 morning ledgers.
• Two candidates with zero catalog hits were assessed and rejected on class: Casefile (agent memory, already covered) and Muumuu Domain (regional consumer domain registrar).

Related

• External MCP Server Catalog - the curated catalog
• MCP Ecosystem Sweeps - all sweep reports