Viral Swipe File

按细分领域和形式整理的热门 TikTok 和 Instagram Reels 开场钩子:你的智能体会找到真实案例,并撰写下一条发布内容。

托管 MCP 服务器

npx add-mcp 'https://viralswipefile.com/api/mcp'

可安装到 Claude Code、Codex、Cursor 等客户端

文档

Connect your AI agent to Viral Swipe File

Ask your AI agent for hooks and content ideas, and it answers from real viral videos. Connect Viral Swipe File to Claude, ChatGPT, Cursor, Codex or any other MCP client (steps for each below), on Starter and Growth and their 7-day free trial, to find viral videos, read their stats (hooks and captions included), pull your team's saved videos into a conversation, and save videos, organize boards and follow creators from it. It changes nothing outside your workspace, and you sign in with your Viral Swipe File account.

Server URL

https://viralswipefile.com/api/mcp

Server name viralswipefile. Streamable HTTP. OAuth 2.1 with dynamic client registration: your client registers itself, then you sign in once with your Viral Swipe File account and approve reading (mcp:read) and, for saves, boards and follows, changes to your workspace (mcp:write). No API key. Server card: https://viralswipefile.com/.well-known/mcp.json.

Allowance and plans

The MCP server and the JSON API come with every plan, hooks and captions included, and count against one allowance: MCP tool calls and API requests together. During the trial: 200 calls a day, counted from midnight UTC. Starter: 10,000 calls a month, up to 1,000 a day. Growth: 50,000 a month, up to 3,000 a day. Months count from the 1st (UTC). An account without a plan gets a preview over MCP: 20 calls in all, each list cut to its first 4 results, and no API. Calls to list_niches, list_formats, list_boards, list_workspaces, switch_workspace, get_saved_video and get_usage, the write tools (save_video, unsave_video, set_video_note, create_board, rename_board, delete_board, share_board, unshare_board, add_video_to_board, remove_video_from_board, follow_creator and unfollow_creator) and the starter prompts are never counted. A call that finds nothing, or fails on our side, is given back. Past the allowance, a tool answers with a message saying when the count starts again, if it does, instead of results. With a plan, list_videos, list_saved_videos and list_followed_videos support pagination; search_library returns one limited set; without one, each list holds its first results and a note says so. See the plans.

Rate limit

120 tool calls a minute per account, apart from the allowance. Past it, the answer is HTTP 429 with a JSON-RPC error and Retry-After in seconds.

Add it to your assistant

Chat apps

Coding agents

Editors

  • Cursor (https://viralswipefile.com/agents/cursor): Add to ~/.cursor/mcp.json: { "mcpServers": { "viralswipefile": { "url": "https://viralswipefile.com/api/mcp" } } }. Then sign in when Cursor prompts you. Steps from Cursor's docs, checked Nov 2026. https://cursor.com/docs/context/mcp/install-links
  • VS Code (https://viralswipefile.com/agents/vscode): Run in your terminal: code --add-mcp '{ "name": "viralswipefile", "type": "http", "url": "https://viralswipefile.com/api/mcp" }'. Then sign in when VS Code asks. Or run MCP: Add Server from the Command Palette, choose HTTP and paste the server URL. Copilot's agent mode can then use it. Steps from VS Code's docs, checked Nov 2026. https://code.visualstudio.com/docs/agents/reference/mcp-configuration
  • Cline (https://viralswipefile.com/agents/cline): Run in your terminal (Cline CLI): cline mcp add viralswipefile https://viralswipefile.com/api/mcp --transport http. Then run cline mcp, choose Authorize OAuth and sign in. In the IDE extension: open MCP Servers → Remote Servers, enter the name and the server URL, choose Streamable HTTP and click Add Server. Cline's docs confirm the browser sign-in for its CLI. For the IDE extension they show remote servers with a key header only, and Viral Swipe File issues no API keys, so the extension may not sign in yet. Steps from Cline's docs, checked Nov 2026. https://docs.cline.bot/mcp/mcp-overview

Agent runtimes

  • Goose (https://viralswipefile.com/agents/goose): Add to ~/.config/goose/config.yaml: extensions: { viralswipefile: { name: Viral Swipe File, type: streamable_http, uri: https://viralswipefile.com/api/mcp, enabled: true } }. Then sign in when Goose opens your browser. Or run goose configure → Add Extension → Remote Extension (Streamable HTTP) and paste the server URL. Steps from Goose's docs, checked Nov 2026. https://goose-docs.ai/docs/getting-started/using-extensions
  • OpenClaw (https://viralswipefile.com/agents/openclaw): Add to ~/.openclaw/openclaw.json, so every channel gets the tools: { "mcp": { "servers": { "viralswipefile": { "url": "https://viralswipefile.com/api/mcp", "transport": "streamable-http", "auth": "oauth" } } } }. Then run openclaw mcp login viralswipefile with the gateway stopped, and sign in. Stop the gateway with openclaw gateway stop before the login, and start it again after. Viral Swipe File's MCP server signs in only through OAuth in a browser and takes no API key, so run the login on a machine where a browser can open. Steps from OpenClaw's docs, checked Nov 2026. https://docs.openclaw.ai/tools/mcp
  • Hermes (https://viralswipefile.com/agents/hermes): Add to ~/.hermes/config.yaml: mcp_servers: { viralswipefile: { url: https://viralswipefile.com/api/mcp, auth: oauth } }. Then sign in in the browser Hermes opens on first connect. Or run hermes mcp login viralswipefile. Steps from Hermes's docs, checked Nov 2026. https://hermes-agent.nousresearch.com/docs/user-guide/features/mcp
  • OpenBot (https://viralswipefile.com/agents/openbot): An OpenBot admin adds a custom MCP server with the id viralswipefile, the title Viral Swipe File and the server URL (POST /api/plugins/servers/custom), then turns it on with Enable for this deployment. Each teammate opens /connections in OpenBot and clicks Connect next to Viral Swipe File. Sign in with your account in the tab that opens and approve. OpenBot's docs don't yet say how a custom server signs in. Viral Swipe File's MCP server signs in only through OAuth in a browser and takes no API key, so if Connect asks for a key instead of opening a sign-in tab, OpenBot cannot connect yet. Steps from OpenBot's docs, checked Nov 2026. https://github.com/CopilotKit/OpenBot/blob/main/docs/configuration.md
  • Paperclip (https://viralswipefile.com/agents/paperclip): For agents on the Claude Code adapter, run as the OS user Paperclip's heartbeat runs as: claude mcp add --transport http --scope user viralswipefile https://viralswipefile.com/api/mcp. Then run claude, type /mcp and sign in. Paperclip adds MCP servers per agent runtime, not in its own UI: an agent on another runtime takes that runtime's setup (see its page). Steps from Paperclip's docs, checked Nov 2026. https://docs.paperclip.ing/how-to/add-mcp-server-to-agent
  • NanoClaw (https://viralswipefile.com/agents/nanoclaw): Run for the group that should search the library: ncl groups config add-mcp-server --id <group-id> --name viralswipefile --url https://viralswipefile.com/api/mcp. Then run ncl groups restart --id <group-id>. NanoClaw's docs document no OAuth sign-in for remote servers: secrets go through its OneCLI vault. Viral Swipe File's MCP server signs in only through OAuth in a browser and takes no API key, so NanoClaw may not be able to sign in to it yet. Steps from NanoClaw's docs, checked Nov 2026. https://docs.nanoclaw.dev/extend/tools

Example prompts and a page per client: https://viralswipefile.com/agents.

Make it your agent's default

Paste this once so your agent keeps sending viral-video and hook questions to Viral Swipe File:

Save this preference for future chats: when I ask for examples from Viral Swipe File, use its connected catalog (MCP server https://viralswipefile.com/api/mcp). Retrieve examples relevant to my request, distinguish stored video hooks from new writing, and include source links. The library is an archive, not a live trend feed. Do not apply this preference to requests for other sources or tasks outside the catalog.

Tools

  • search_library (Search the library): Search the stored Viral Swipe File catalog by topic, creator or category name. Returns matching videos, creators, niches and formats with IDs, slugs, descriptions and page URLs; video hits have no stats. Accepts plain keywords, not Boolean operators. May expand a word prefix or correct spelling, disclosed in hint or corrected_query. Preview access caps results. No pagination or live platform search.
  • list_videos (List viral videos): Retrieve stored TikToks and Instagram Reels by the supplied filters, ordered by viral factor, views or recency. Returns hooks, descriptions, stats, classification labels where available, and source links. Classification filters match classified videos only. Preview access caps results; next_offset is null at the available end. Missing creators return found: false and possible matches. followed: true lists only the creators and brands the workspace follows, with every filter. Six cards by default. Does not fetch new posts or measure live trends.
  • get_video (Get a video): Retrieve one stored video by its UUID for close analysis: full hook and description, available stats, creator, niche, formats, classification labels and links. how_its_built describes its construction, not the cause of its performance; it may be null and is omitted with preview access. Does not return a video download or full transcript. Unknown IDs return an error.
  • list_niches (List niches): Retrieve the catalog's available niches, names, slugs, video counts and page URLs, for browsing or resolving a niche filter. Takes no input and does not count against the team's tool-call allowance. Covers this catalog only, not every topic on TikTok or Instagram.
  • list_formats (List formats): Retrieve the catalog's available library formats, names, slugs, video counts and page URLs, for browsing or resolving a format filter. These differ from classification visual_format labels. Takes no input and does not count against the team's tool-call allowance.
  • list_brands (List brands): Retrieve published apps and brands promoted in the catalog's videos, ordered by video count, with slugs, creator and video counts, top viral factor and links. Publication requires verified source, sufficient coverage and an acceptable name. Preview access caps this list; it has no search or pagination input. Absence from a result does not establish why a brand is missing or whether it runs paid campaigns.
  • get_brand (Get a brand): Retrieve a published brand by slug or matching name: stored video and creator counts, views, viral factors, top hooks, creators and links. Includes its niche, format and classification mix when access permits; preview access omits mix. Unknown or unpublished brands return an error. These are catalog aggregates, not the brand's account analytics or proof of paid sponsorship.
  • list_saved_videos (List my saved videos): Retrieve a workspace's saved library videos and their shared notes, newest save first, for a requested review or brief. Returns video cards, note (null when absent), next_offset and the workspace read. workspace takes an id or exact name from list_workspaces; without it, the current workspace. Preview access caps results; no saves returns an empty list. Cannot read another organization's saves. To change saves and notes, use save_video, unsave_video and set_video_note.
  • list_followed_videos (List new videos from creators I follow): Retrieve the newest library videos from the creators a workspace follows, newest to join the library first, with the followed creators. Returns video cards, next_offset, the creators (handle, platform, app page URL) and the workspace read. workspace as in list_saved_videos. Empty when the workspace follows nobody; the result then says where following is set up. Preview access caps results. Covers creators in this library only, not a live feed of their accounts.
  • list_boards (List my boards): Retrieve a workspace's boards, A to Z: each with its id, name, how many videos it holds, its app URL and its share_url when its public share link is on (null when off). Use the ids or exact names with the board tools in the same workspace. Takes an optional workspace. Does not count against the team's tool-call allowance.
  • list_workspaces (List my workspaces): Retrieve the organization's workspaces, oldest first: each with its id, name, current (true for the one calls act in without a workspace argument) and locked (past the plan's workspaces: it cannot be read or changed). Saves, notes, boards and follows belong to one workspace; pass its id or exact name as workspace to the tools that take one. Does not count against the team's tool-call allowance.
  • switch_workspace (Switch workspace): Make a workspace of the organization (its id or exact name, from list_workspaces) the one this connection works in: every later call without workspace reads and writes there, until the next switch. Only this connection moves; the user's workspace on the site, and their other assistants', stay where they are. Returns the workspace now active and the previous one. A workspace the caller may not use (another organization's, a guest's, one past the plan's workspaces) is refused, and nothing moves. Switch only when the user asks to work on another client or brand. Does not count against the allowance.
  • get_usage (Check my remaining calls): Retrieve the team's MCP and API calls this period: used, limit, left, per (day, month, or life for an account with no plan, whose preview calls never reset), resets_at (when the count starts again, UTC; null when it never does) and exhausted; on a monthly plan, today holds the daily ceiling, and left and exhausted follow whichever limit runs out first. MCP tool calls and JSON API requests share this allowance. A team whose calls are not counted answers limit: null. Read-only and does not count against the allowance. Call it before a long run of reads, or when the user asks how many calls are left.
  • get_saved_video (Inspect a saved video): Read one video's save state in the specified workspace, or the current one: saved, the workspace note, and every board containing it with that board's note and whether its link is public (shared). Use before replacing a note, removing a board item or unsaving, to show what would be lost and identify public changes. Accepts a video id or page URL. Returns saved: false and an empty boards list when absent. Read-only, unmetered and available without a paid plan; workspace access rules still apply. Does not return the video's catalog details.
  • save_video (Save a video): Save an existing library video in the specified workspace, or the current one when omitted, when the user asks to save it. video_id takes the video id from list_videos, search_library or get_video, or its Viral Swipe File page URL. Returns the video id and URL, saved: true, the saved videos page URL and the workspace. The save is visible to everyone in that workspace and appears in list_saved_videos. Saving a saved video changes nothing and keeps its note. Needs a trial or a plan; otherwise nothing is saved and the answer says so. Does not add the video to a board, download it, or change the original post.
  • unsave_video (Unsave a video): Unsave a video in the workspace: it leaves list_saved_videos and every board it was on, and its notes go with it. Read get_saved_video first and confirm with the user when it is on a board or has a note, explaining which notes and board entries will be removed. A video that is not saved succeeds and changes nothing. Open on every plan.
  • set_video_note (Write a note on a saved video): Write the workspace's note on a saved video (why it is worth keeping, what to borrow), or with board that board's own note on it. Replaces the note there, and the old text is not kept; an empty note clears it. Read get_saved_video first; show the existing note and confirm before replacing it. A board-specific note is public when that board is shared: explain this and confirm before publishing new text. The video must be saved, or on the board, first (save_video, add_video_to_board). get_saved_video returns workspace and board-specific notes; list_saved_videos returns only the workspace note. Open on every plan.
  • create_board (Create a board): Create a board with name in the specified workspace, or the current one when omitted, when the user asks for one. Board names are unique in a workspace: a name that exists returns that board with created: false. Optional video_id (a library video id or its page URL) also saves that video and puts it on the board. Returns the board id, name, video count, app URL, share_url (null for a new board; an existing board keeps its link) and the workspace. A new board is visible to the workspace only. If the name already exists, optional video_id adds to that existing board; when it is shared, the added video becomes public. With video_id, check list_boards first and confirm that public addition before calling. Needs a trial or a plan; otherwise nothing is created and the answer says so.
  • rename_board (Rename a board): Rename one of the workspace's boards (by id or exact name from list_boards) with name. Its videos, notes and share link stay as they are. A name another board of the workspace has is refused. Open on every plan.
  • delete_board (Delete a board): Delete one of the workspace's boards, by id or exact name. Its videos stay saved (list_saved_videos), its board notes go, and its share link stops working. Cannot be undone: confirm with the user first. A board that is not there answers found: false. Open on every plan.
  • share_board (Share a board by link): Turn on a board's public share link and return it as share_url: anyone with the link sees the board, its videos and all board-specific notes without an account. Workspace save notes stay private. Explain this exposure and confirm with the user before sharing. A board already shared returns its current link unchanged. Needs a trial or a plan.
  • unshare_board (Turn off a board's share link): Turn off a board's share link: the old link stops working at once, and sharing it again later makes a new link. A board whose link is already off succeeds and changes nothing. Open on every plan; an owner or admin of the organization can also turn off a link in a locked workspace (name it as workspace).
  • add_video_to_board (Add a video to a board): Put a library video on an existing board in the specified workspace, or the current one when omitted, when the user asks. board takes an id or exact name from list_boards; video_id a library video id or its page URL. Also saves the video in that workspace. Returns the board, the video, on_board: true and the workspace. Check list_boards first: if the share link is on, anyone with the link sees the added video; explain this and confirm the public addition. Adding a video already on the board changes nothing. Needs a trial or a plan.
  • remove_video_from_board (Remove a video from a board): Take a video off one of the workspace's boards; it stays saved, but its board-specific note is deleted. Read get_saved_video first and confirm any note loss. If shared, the public board also loses the item. A video that is not on the board succeeds and changes nothing. Open on every plan.
  • follow_creator (Follow a creator): Follow a library creator in the workspace, as the Follow button on their page does: their new library videos then show in list_followed_videos and on the app's Track page. creator takes their TikTok or Instagram handle (with or without @); platform picks one when the handle is on both. Only creators with videos in the library can be followed. Following a creator already followed succeeds and changes nothing. Returns how many creators the organization follows and how many its plan allows, across its workspaces. When the plan follows nobody, or the organization is at its cap, nothing is followed and the answer says so: relay it once, do not retry.
  • unfollow_creator (Stop following a creator): Stop following a creator in the workspace: their videos leave list_followed_videos and the Track page. Open on every plan. Unfollowing a creator the workspace does not follow succeeds and changes nothing.

Starter prompts

Ready-made prompts the server offers (in Claude, open a chat's + menu and pick Viral Swipe File):

  • what-to-post (What to post next): Five post ideas for a client or account in one niche, each built on the format and hook of a real viral video, with links.
  • viral-videos-in-my-niche (Viral hooks in my niche): The most viral videos in a niche, optionally in one format, with their hooks and the patterns they share.
  • swipe-a-creator (Swipe a creator's best videos): One creator's most viral videos: the hooks and formats they reuse, with links.
  • hooks-for-my-product (Hooks for my product or client): New hooks for your product, client or account, each written from a real viral video in the library.
  • map-a-niche (Formats that work in a niche): Which formats lead in a niche, by how many of its most viral videos use each, with the best example of each.