Suger
Cloud marketplace and go-to-market data in chat
ChatGPT · Claude · 137 tools
AI Connectors Directory
We catalogued the ChatGPT apps directory and the Claude connectors directory in August 2026, 2,305 listings on one side and 1,253 on the other, and merged them into one catalogue of 3,150 products. Each profile shows where the product is listed and what each marketplace listing declares, with live-endpoint measurements where the server answered.
Across marketplaces
2,289 products are listed in the ChatGPT apps directory and 1,253 in the Claude connectors directory, but just 392 appear in both, matched by publisher domain and product name, and merged into one profile here. Since one MCP server can serve both surfaces, the other 2,758 are a distribution decision away from doubling their reach.
Cloud marketplace and go-to-market data in chat
ChatGPT · Claude · 137 tools
Sync Teamwork projects and tasks into Codex
ChatGPT · Claude · 100 tools
Global hiring cost comparisons and take-home pay estimates
ChatGPT · Claude · 95 tools
Configure AI engagement agents and WhatsApp templates
ChatGPT · Claude · 95 tools
Cooperative contract and supplier search for public agencies
ChatGPT · Claude · 89 tools
Course recommendations for data and AI skills
ChatGPT · Claude · 87 tools
Music industry data and reporting
ChatGPT · Claude · 86 tools
Send and inspect transactional email from ChatGPT
ChatGPT · Claude · 85 tools
Customer evidence search for product decisions
ChatGPT · Claude · 80 tools
Ambassador and creator program management in chat
ChatGPT · Claude · 78 tools
Marketing platform data and campaign creation from chat
ChatGPT · Claude · 77 tools
Subscription revenue analytics queried in plain language
ChatGPT · Claude · 76 tools
Browse
Each hub lists every connector filed under it across both marketplaces, with the bucket's marketplace split and median tool depth.
The default bucket on both marketplaces.
Browse ProductivityThe largest of the regulated buckets, unusually well represented for its sector, and with a higher-than-average share of partner verification.
Browse FinanceWhatever the marketplaces could not place: the most heterogeneous group in the catalogue, and a useful reminder that category fit matters less than whether the underlying data is worth asking about..
Browse OtherThe broadest of the business buckets, mixing horizontal workflow tools with narrow single-purpose utilities.
Browse Business OperationsWarehouses, BI platforms, semantic layers, and a long tail of dataset-specific lookup services, mostly built by the platform vendors themselves.
Browse Data & AnalyticsAirlines, hotels, rail, OTAs, and destination services, clustered tightly around search and inspiration rather than booking.
Browse TravelInfrastructure, CI, observability, source control, and API platforms: the bucket with the highest concentration of first-party servers built by the platform vendor itself, and the deepest tool surfaces in the directory: one tool per resource per verb pushes counts well above the median..
Browse Developer ToolsListings split between study aids aimed at individual learners and research tools that search a specific corpus.
Browse Education & ResearchSmall relative to the size of the sector, which reflects how carefully the category has to be entered.
Browse Healthcare & Life SciencesStreaming, publishing, gaming, sports, and fan platforms.
Browse Entertainment & MediaCRM, outbound, enrichment, advertising, SEO, and campaign platforms, with enrichment and lookup services making up a large share.
Browse Sales & MarketingGeneration and editing utilities dominate.
Browse Creative & DesignMessaging, email, telephony, meetings, and contact-centre platforms.
Browse CommunicationStorefronts, marketplaces, logistics, and post-purchase services, small relative to the size of the sector.
Browse Commerce & ShoppingCase law, contract, compliance, and legal-research services: small, and almost entirely retrieval-shaped.
Browse LegalOne of the smallest buckets, and conspicuously conservative: most listings describe themselves as read-only and route anything consequential back to the vendor’s authenticated console, which is the right pattern for the category..
Browse SecurityThe smallest bucket in the directory by a wide margin, almost entirely retrieval over a public or member dataset.
Browse NonprofitCategories come from the marketplace listings themselves, mapped onto one shared set, and at row level they are noisy: connectors are filed by their publisher, and plenty sit under a heading that does not describe the product well. Treat the grouping as a way to navigate the directory, not as a market segmentation.
Notable
Everything else in the Claude directory is published by a third party, either as a verified partner or community-listed. The ChatGPT directory's equivalent: 34 of its apps are OpenAI-built.
Public data on how Claude usage varies by country and occupation
9 tools · Data & Analytics
ICD-10-CM and ICD-10-PCS lookup, search and validation
9 tools · Healthcare & Life Sciences
Preprint search across bioRxiv and medRxiv
8 tools · Healthcare & Life Sciences
Medicare Part B coverage determinations from the CMS database
8 tools · Healthcare & Life Sciences
Work SharePoint, OneDrive, Outlook, and Teams data inside Claude
7 tools · Productivity
Search PubMed citations and retrieve PubMed Central full text
7 tools · Healthcare & Life Sciences
Bioactivity, targets and mechanism data from the ChEMBL database
6 tools · Healthcare & Life Sciences
Search ClinicalTrials.gov by condition, sponsor, investigator or eligibility
6 tools · Healthcare & Life Sciences
Look up and validate US healthcare provider NPI numbers
3 tools · Healthcare & Life Sciences
Who publishes
And 95.6% of ChatGPT developers do too. Both catalogues are nearly all tail: a company with a well-built second or third connector is still an unusual thing.
Out of 3,150. The overlap is where the strongest publishers cluster: the companies that treated distribution as part of the build, not an afterthought.
Privacy policy and terms of service are effectively table stakes in the ChatGPT review process, and 99.7% of listings authenticate at install.
The rest publish documentation but no endpoint URL in the directory record. They are real listings; the URL has to come from the publisher's own setup docs.
Method
Both marketplace directories were walked in August 2026: 2,305 ChatGPT listings and 1,253 Claude listings, nothing dropped. Products listed in both were merged into one profile when the publisher domain and product name match; where the domains disagree but the product is plainly the same, the profile says the match is by name.
Where a Claude listing publishes an endpoint URL, Node8 called it directly, an MCP initialize followed by tools/list and nothing more, and recorded what came back: a tool list, a credential gate, or a CDN block. Everything comparative on these pages (percentiles, category ranks, tier rarity, marketplace overlap) is computed by Node8 from the snapshot. Publisher descriptions are quoted and attributed, never presented as ours. The full inventory of tools the live servers returned is at verified MCP tools, and the analysis of what the measurements show is in The State of Claude Connectors.
Counts are accurate as of August 2026 and will drift as the directories change. Listing a company here is not an endorsement, and Node8 has no commercial relationship with these companies unless separately stated.
FAQ
An AI connector is an MCP server a company publishes so an assistant (ChatGPT, Claude, Microsoft Copilot) can reach that company’s data or take actions against it in conversation. The server declares a set of tools; the assistant calls them. ChatGPT apps and Claude connectors are the same idea distributed through two different marketplaces, and one server can back both.
Both marketplace directories were captured in August 2026: 2,305 listings in the ChatGPT apps directory and 1,253 in the Claude connectors directory. Products listed in both are merged into one page, matched by publisher domain and product name; 392 of the 3,150 products here are on both marketplaces. The comparisons on these pages are computed by Node8.
It is the number of MCP tools a connector declares in its Claude listing (the ChatGPT directory does not publish counts). The Claude-directory median is 11; the mean is 19.7, pulled up by a long tail that reaches 255. Node8 also called every published endpoint directly: 222 answered, and 31% of those publish a count that does not match their server.
392 of the 3,150 products in this directory are listed in both the ChatGPT and Claude directories, matched by publisher domain and product name. The other 2,758 are on exactly one marketplace, which is a distribution decision rather than a technical one: the same MCP server can serve both.
Community-listed means self-published; partner verification is a review step; and 9 connectors were built by Anthropic itself. The tier says who published and whether it was reviewed, not whether the connector is good. The ChatGPT directory has no equivalent tier, but 99.7% of its listings authenticate at install.
Categories come from the marketplace listings themselves, mapped onto one shared set, and at row level they are noisy: connectors are filed by their publisher, and plenty sit under a heading that does not describe the product well. Treat the grouping as a way to navigate the directory, not as a market segmentation.
No. This is an independent catalogue of publicly listed connectors and apps, published as reference. Node8 builds MCP servers for its own clients and has no commercial relationship with the companies listed here unless separately stated.
Node8 designs the tool surface, builds and secures the MCP server, and takes it through review on every marketplace that matters. Most first connectors ship in eight to twelve weeks.