Give your AI assistant secure, plain-language access to Unspam: spam checks, inbox placement, screenshots, heatmaps, and recurring autopilot tests. Connected over OAuth, with no API keys to copy.
The server speaks JSON-RPC 2.0 over HTTP and authenticates with OAuth 2.1 + PKCE. Your assistant opens a browser, you sign in once, and the client caches the token. Nothing to paste, nothing to rotate by hand.
Around 30 tools cover spam checks, inbox placement, autopilot (scheduled tests), screenshots and heatmaps, plus account status. Your assistant reads the catalog and picks the right call for each request.
Claude Desktop (custom connector or the mcp-remote bridge), Claude Code, Cursor, ChatGPT Developer Mode (Plus, Pro, Business), Windsurf, and any MCP client that supports remote streamable-HTTP servers.
The link shows up under Connected apps in your Unspam account, so you can revoke it whenever you want. MCP access is available on paid plans only, gated to your subscription quota.
Approve one OAuth screen, hand your assistant the full Unspam tool set, then ask for tests in plain language.
Your client registers itself with Dynamic Client Registration, then runs the Authorization Code grant with PKCE (S256). You sign in through the browser and approve a single consent screen scoped to mcp.
After that, the client caches the token and renews it for you. A missing or invalid token returns 401, the wrong scope returns 403, so the failure modes are explicit.
Once connected, your assistant sees every Unspam tool: start_spam_check, get_spam_check_result, get_spam_check_screenshots, get_spam_check_heatmap, the inbox-placement set, the scheduled-test (autopilot) set, and get_account_status for limits.
No separate SDK to wire up. The client reads the catalog over MCP and exposes each tool to the model directly.
Say "Run a spam check and tell me when it is done" or "Start an inbox placement test across all mailboxes" and the assistant sequences the tool calls, polls for the result, and reports back.
Set up recurring coverage the same way: "Schedule a weekly automatic test and email me if anything hits spam." The autopilot tools create the sender, the schedule, and the notification for you.
MCP needs a paid plan with API access. Add the server URL to your client, approve the OAuth screen, then talk to your assistant in plain language.
MCP needs API access, which is included on all paid plans. Free accounts cannot connect. Check your subscription and monthly limits anytime with get_account_status.
Claude Code: claude mcp add --transport http unspam https://api.unspam.email/mcp. Cursor: add {"url":"https://api.unspam.email/mcp"} to ~/.cursor/mcp.json. Claude Desktop and ChatGPT: paste the same URL as a custom connector.
Your client opens a browser, you sign in to Unspam, and you approve access scoped to mcp. The token is cached, so you only do this once per client.
Try "Run a spam check and tell me when it is done." The assistant calls start_spam_check, polls get_spam_check_result, and reports the score back in the chat.
Supported clients, pricing, authentication, rate limits, and how to revoke access.
mcp-remote bridge), Claude Code, Cursor, ChatGPT in Developer Mode (Plus, Pro, Business), and Windsurf. The setup is always the same server URL: https://api.unspam.email/mcp.get_account_status from your assistant to see your subscription and remaining monthly limits.mcp. You sign in through a browser once and the client caches the token. There are no API keys to copy or rotate by hand, and you can revoke the connection anytime.get_account_status. If you exceed your limit, the server returns 429. A missing or invalid token returns 401 and the wrong scope returns 403.