🎉 Get 7 months FREE on any annual plan — no risk, cancel anytime
06
HR
00
MIN
00
SEC
https://mcp.cudekai.com
Humanizes AI-generated text through humanize_text, stripping the giveaway patterns that make writing feel machine-made and lifting the human score in a single pass. Turns an AI-sounding cold email or blog draft into a polished version ready to send, without a second round of manual editing.
Analyses any passage through detect_ai_text and returns an AI probability score, confidence score, and human/AI mixed classification with a sentence-level breakdown. Runs cleanly as a pre-publish check, and when the score is too high, the assistant loops straight into humanize_text to bring it down before the article ships.
Scans text through check_plagiarism against public sources, producing a similarity score, matched sources, and a full plagiarism report. Scales well for reviewing student essays or submissions at volume; copied passages surface with the URL they matched against, so originality checks happen well before grading.
Checks every reference in a document through check_citations to catch fake citations, hallucinated citations, and inaccurate sources. Runs a full citation audit across research papers, flagging fabricated references with the reason each one failed source verification — no more chasing down phantom footnotes by hand.
Spots AI-generated images from Midjourney, DALL-E, Stable Diffusion, and other models through detect_ai_image, including deepfakes and synthetic imagery. Batch-checks folders of uploaded images in one prompt, returning an authenticity score for every file before anything gets forwarded, trusted, or published.
Extracts a full timestamped transcript from any YouTube URL through get_youtube_transcript, drawing from auto-generated captions or existing video subtitles. Converts a ninety-minute lecture or podcast into clean, ad-free segments the assistant can summarize, quote, or pull key points from in the same exchange.
Flags AI-generated videos, deepfakes, Sora-style clips, and face-swap manipulations through ai_video_detection across uploaded or linked media. Pairs with detect_ai_image for a single-chat workflow that lets newsrooms, verification teams, and researchers vet media authenticity before anything gets acted on.
Runs several CudekAI tools in sequence from one prompt: detects a draft, humanizes anything over the threshold, and then plagiarism-checks the finished version. Handles the orchestration underneath so the user only has to describe the outcome they want.
Open Claude Desktop or Claude.ai, go to the connectors panel, and add a custom connector using the CudekAI remote MCP server URL. Approve the consent screen and toggle the connector on to unlock every CudekAI tool inside every new chat.
Open cursor settings, go to the MCP section, and add a new MCP server entry pointing at the CudekAI connector URL. Save the entry, and Cursor picks up every CudekAI tool for direct use inside the IDE.
Add CudekAI to Codex CLI with a one-command install directly from the terminal. Handles the OAuth handshake automatically and registers every tool for use in Codex sessions.
Configure other MCP clients through manual setup using HTTP transport, streamable HTTP, or mcp-remote. Connect any MCP-compatible assistant with a custom MCP setup for flexible integration.
Pastes the CudekAI connector URL into the client and completes OAuth sign-in through the browser. Persists the connection across every future session with no reconnect required.
Reads the natural-language request and picks the right CudekAI tool for the task automatically. Removes the need for command syntax or manual tool selection.
Returns every score, output, or best result directly inside the chat window. Delivers structured replies the assistant can quote, format, or feed into the next step.
Handles humanization and detection tasks across 40+ languages through built-in language code detection. Processes non-English content natively and translates before humanizing when the workflow calls for it.
Accepts a tone hint for professional, casual, academic, or creative output on every humanized request. Adapts phrasing and word choice to match the style without a second rewrite pass.
Returns multiple humanized variations of the same input, each with a per-variation score. Surfaces the strongest rewrite through best-of scoring while keeping alternatives available for review.
Reports remaining requests, usage this cycle, and available balance through the get_account tool. Runs as a read-only lookup that never spends quota.
Handles bulk content through batch processing, running several passages through the same tool in one session. Scales cleanly across long documents and volume reviews section by section.
Returns machine-readable JSON that the assistant can parse, filter, and feed into downstream steps. Makes chaining, scripting, and formatting reliable across every MCP client.
Runs every check and rewrite inside the chat window, removing the paste-in, paste-out loop that separate apps require. Keeps the workflow frictionless from first prompt to final result.
Text-based tools accept a text parameter with a minimum character count and a per-request word cap that varies by plan tier. URL-based tools like get_youtube_transcript accept a video URL or video ID as supported input.
Detection tools return numeric scores on a percentage range from 0 to 100 alongside a JSON structure with a per-sentence breakdown. Transcript output supports SRT, VTT, and plain text with timestamped segments in every response.
CudekAI MCP applies a short-window rate limiter of 30 calls per minute per client alongside a per-cycle allowance tied to the plan. Requests beyond the window return a cooldown notice with a retry-after value.
CudekAI MCP is included on every active paid subscription. The free plan does not include MCP access, so an upgrade to a paid tier is required before the connector will authenticate.
The Pro plan, Ultra plan, and API plan each include MCP access with their own monthly request allowance, per-request word cap, and variations per plan. Overflow requests fall back to the API balance where available, with top-up credits and a shared budget across CudekAI tools on higher tiers.
CudekAI MCP helps troubleshoot connection failures, rate limits, quota exhaustion, and upgrade requirements. It reconnects clients or disconnects and removes connector entries when per-request limits are exceeded. CudekAI simplifies troubleshooting to help restore a working connection.
Connect now by copying the connector URL to get started in minutes. It lets users upgrade to connect and work without switching tabs. CudekAI makes connecting simple for a smoother workflow.
https://mcp.cudekai.com