Claude connector for SaaS:what it takes
Updated · 4 min read
If your users work in Claude, a connector puts your product inside that work: Claude reads their data and takes actions on their account. This is what Anthropic requires from such a connector, from sign-in to the directory listing.
What a connector gives you
A connector is a remote MCP server added to Claude. For a SaaS product that means:
- One sign-in for every app: claude.ai, the desktop and mobile apps, Cowork and Claude Code.
- On Team and Enterprise plans an organization Owner enables the connector for the whole team.
- A directory listing is optional. Users can add the connector by its server URL; the directory adds visibility and a quality label.
- The same server works as a ChatGPT plugin, with its own sign-in configuration and listing.
Sign-in: OAuth, not API keys
When tools act on a user's account, every user signs in with their own account through OAuth 2.0. Claude's OAuth client has a few requirements that setups most often trip over:
- Client registration: by default through Dynamic Client Registration or a Client ID Metadata Document. With heavy directory traffic CIMD is better, because DCR registers a new client on every fresh connection.
- On request (by email to mcp-review@anthropic.com) Anthropic can hold the credentials of an OAuth client you create for Claude.
- PKCE with the S256 method on every authorization. A client_credentials grant with no user involved isn't supported.
- A 401 response with a WWW-Authenticate header pointing to your resource metadata. Claude won't start sign-in from a 200 response.
- The redirect URI https://claude.ai/api/mcp/auth_callback, plus localhost redirects on any port for Claude Code.
- OAuth endpoints must answer within 10 seconds, token refresh within 30.
Some tools without sign-in
If some tools work without an account (searching public docs, for example), Claude supports lazy authentication: users use those right away and sign in only when they reach a tool that needs their account. For companies with SSO there's Enterprise Managed Auth, which connects users through the organization's identity provider without a consent screen.
A fixed API key in a request header, entered once by an organization Owner, is in beta and only available to some organizations. Don't plan a connector for everyone around it.
Tools that pass review
- Separate tools for reading and writing. A single catch-all api_request tool with a method parameter is rejected.
- Every tool has a title and a readOnlyHint or destructiveHint annotation. Read-only tools can run without asking, tools that change data always ask for approval.
- Names of up to 64 characters and descriptions that say exactly what the tool does. No instructions steering Claude's behavior and no promotion.
- Specific error messages instead of "Internal Server Error", and reasonably sized responses.
- Your own API on your service's domain. Connectors that move money or generate images, video or audio with AI models aren't accepted.
A separate URL for every customer
In many SaaS products each customer has their own subdomain. In the submission you choose one shared URL, a fixed list of URLs (one per region, for example) or a pattern the user's URL has to match. With a pattern Anthropic recommends CIMD, and such listings take longer to review.
Submitting to the Connectors Directory
Anyone on a paid Claude plan can submit, through the portal at claude.ai/directory/manage. Besides the server you need a fully populated test account, public documentation (a blog post or help-center article is enough) and seven policy acknowledgments. The full checklist for both directories is in How to get listed in ChatGPT and Claude.
Community or Verified
Anthropic scans each submission automatically for policy compliance and lists it as a Community connector by default. Listings the system flags as especially useful go on to Verified review, where reviewers test every tool. There's no separate application for that status. Verified connectors get a checkmark next to their name, and before someone connects a Community connector Claude reminds them it hasn't been reviewed in depth.
- Do I need a Team or Enterprise plan to submit?
- No. Anyone on a paid Claude plan can submit.
- Can customers connect before the directory listing is live?
- Yes. Anyone can add the connector by its server URL as a custom connector. The directory adds visibility and a label; it doesn't change how the tools work.
- Do I need a separate OAuth app for every enterprise customer?
- No. Enterprise customers connect through the same OAuth app as everyone else, and what each user can reach is set by their permissions in your product.
- MCP integration: your ERP, store and CRM in ChatGPT and Claude
- Aesthetic clinic websites in Poland
- Online booking for a private clinic
- An AI assistant for your clinic: what's allowed in 2026
- ChatGPT product feed: get your products into ChatGPT shopping
- Clinic advertising in Poland: what a website may say
- Your clinic in ChatGPT and Google
- Patients from abroad: a medical tourism website
- Claude connectors: what they are and how to add one
- MCP servers in ChatGPT: how to connect one
- How to connect WooCommerce to ChatGPT in 2026
- OAuth for MCP servers: sign-in in ChatGPT and Claude
- MCP Apps: interface in ChatGPT and Claude
- MCP server security: what to check before launch
- How to test an MCP server in ChatGPT and Claude
- MCP server or API? When you need both
- Hiring a software studio in Poland
- Your Shopify or WooCommerce store in ChatGPT
- What is an MCP server? Explained for online stores
- How to sell in ChatGPT in 2026
- MCP server, ChatGPT plugin or Claude connector?
- How to get listed in ChatGPT and Claude
Want a Claude connector for your product?
Claude connectors