Dify MCP setup: give your Dify AI agent live TikTok and Instagram data

Updated

Virlo is social listening for TikTok and Instagram Reels. A Dify agent can reason, branch and write, yet none of its built-in tools report how TikTok and Instagram videos are performing this week. Register Virlo as an MCP server in Dify and every agent node, chatflow and workflow in your workspace can query trends, sounds, hooks and outlier videos from a database of 23.1M viral videos. Dify is an open-source platform for building LLM apps, agents and visual workflows, on Dify Cloud or self-hosted.

  • Add Virlo once in the MCP tab and reuse it in every app
  • Sign in through Dify or send a bearer header
  • HTTP Request node for fixed REST pulls into a workflow

Part of Virlo's AI assistant guides. Building the same flow in another canvas? Compare Flowise and n8n.

How do I add the Virlo MCP server to Dify?

Six steps, all inside the Dify console. Virlo becomes a workspace tool provider, so you set it up once and pick it from any app afterwards.

  1. Step 1: Create your Virlo account for free

    Create your free Virlo account in under a minute. Already signed up? Open API Keys.
  2. Step 2: Open the MCP tab in Dify Tools

    In the Dify console, go to Tools (shown as Integrations > Tools on newer builds), switch to the MCP tab and add an MCP server over HTTP. Paste the Virlo server URL below.

    Server URL
    https://dev.virlo.ai/api/mcp/mcp
  3. Step 3: Name it and set the server identifier

    Call the provider Virlo and give it the identifier below. Dify uses the identifier to link your apps to this server, and exported app files reference it too, so pick it once and keep it.

    Server identifier
    virlo
  4. Step 4: Authorize the connection

    Leave Dynamic Client Registration switched on and Dify opens Virlo's sign-in page; approve it and Dify stores the token. Prefer a static key? Add the header below with a key from your Virlo dashboard instead.

    Header (key route)
    Authorization: Bearer virlo_tkn_YOUR_KEY
  5. Step 5: Attach Virlo to an Agent node or Agent app

    In a Chatflow or Workflow, drop in an Agent node and tick the Virlo tools under MCP, or add a single Virlo tool as its own Tool node. Agent apps list the same tools in their Tools panel.

  6. Step 6: Run your first research question

    Open the debug and preview panel and send a question. The agent log shows each Virlo tool it called and what came back.

    Your first prompt
    Use Virlo to find the 5 TikTok sounds trending fastest in [your niche] this week, with how many videos used each one and a sample video link.

Want the same numbers every run instead of an agent deciding? Add an HTTP Request node that calls Virlo's REST API with your key stored as an environment variable, then pipe its JSON into an LLM node for the write-up.

Agent node vs Tool node vs HTTP Request node: where does Virlo go in a Dify app?

Dify building blockHow it uses VirloPick it when
Agent node (Chatflow or Workflow)Chooses among Virlo's research tools from your instruction and chains several callsThe question changes run to run, such as a chat assistant for your content team
Tool nodeCalls one Virlo MCP tool with inputs you wire from earlier nodesYou know exactly which lookup you need, for example a creator check on a handle from a form
HTTP Request nodeCalls a REST endpoint like GET /v1/content-trends with a bearer headerA scheduled workflow that should fetch identical fields every time
Agent appA ready-made chat agent with Virlo enabled in its Tools listYou want a shareable web app with no canvas at all

All four reach the same Virlo account. The MCP provider is added once per workspace; the HTTP route needs only the key.

Dify on its own vs Dify with Virlo: what can the agent answer?

Request to the Dify agentDify aloneDify with Virlo
List this week's rising TikTok soundsNeeds a search plugin and returns articles about musicSounds ranked by how many TikTok videos used them in the last 7 days
Draft five hooks for our launchHooks written from the model's general knowledgeHooks rewritten from real spoken openers on videos that beat their creator's usual views
What format should our Reels follow?A generic best-practice listTrends to copy, each with the part to keep and the part to swap
Is @[competitor] growing on TikTok?No way to checkTheir recent videos, the ones that broke out, and the Virality Score of each

Which TikTok sounds would a Dify agent pull this week?

Ask a Virlo-enabled Dify agent for trending audio and it returns a table like this one, pulled live, ready to hand to an LLM node or a Slack message.

In Virlo's data, 316 TikTok videos used "original sound - hazel gray" in the last 7 days, more than any other sound on TikTok as of November 19, 2026.

As of November 19, 2026 at 12:03 UTC

Which TikTok sounds would a Dify agent pull this week?
SoundArtistTikTok videos (7 days)Licensed for brands
Cover art for original sound - hazel grayoriginal sound - hazel grayhazel gray316Yes
Cover art for Flower In A Hurricane (feat. Tame Impala, SZA, King Gizzard & the Lizard Wizard) [From GTAVI: Original Score]Flower In A Hurricane (feat. Tame Impala, SZA, King Gizzard & the Lizard Wizard) [From GTAVI: Original Score]Justice & Tame Impala & SZA & King Gizzard & the Lizard Wizard287Yes
Cover art for original soundoriginal soundspam_lilah1276Yes
original sound - cluberichcluberich264Yes
original sound - rosyconfessionsᝰ.ᐟ252Yes

See all 300 trending TikTok sounds

Usage history and the TikTok videos behind every sound are in the Virlo app.

Open sounds in Virlo

How this sample is made

Source
Virlo's sound rankings for TikTok, built from the TikTok videos Virlo collects.
Window
TikTok videos posted in the last 7 days that Virlo collected. Counts are TikTok videos in Virlo's data, not every video on the platform.
Method
Ranked by the number of videos using each sound in the window. "Licensed for brands" = the sound is in TikTok's commercial music library.
As of
November 19, 2026 at 12:03 UTC. Rebuilt every day.
Made by
Numbers are pulled automatically from Virlo's data; the text around them is written by the Virlo team.

Suggested citation: Source: Virlo, “Dify MCP setup: give your Dify AI agent live TikTok and Instagram data”, virlo.ai/dify

How do I build a weekly TikTok trend digest workflow in Dify?

A Workflow app with a schedule makes a simple research desk for marketers. One layout that works well:

  • Start with a Schedule trigger set to Monday morning (or a Start node if you run it by hand).
  • Add an Agent node with the Virlo MCP tools and an instruction that names [your niche] and the platforms you care about.
  • Ask the agent for trends to copy, the top outlier videos and the breakout sounds, as structured JSON.
  • Pass that JSON to an LLM node that writes a short brief: three ideas, each tied to a source video.
  • End with an HTTP Request node that posts the brief to your Slack or Discord webhook, or a Template node that formats it for email.

Virality Score tells the agent how far a video beat its creator's usual views. Scores of 18 and above count as strong, so the digest surfaces breakouts from small accounts instead of big names posting average work.

What should a Dify workspace admin know before sharing Virlo?

The MCP provider lives at workspace level, so every app builder in that workspace can attach Virlo tools once it is authorized. If you want research limited to certain apps, keep a separate workspace for them.

The server identifier travels with exported DSL files. When you import an app into another workspace or a self-hosted Dify, add the Virlo server there with the identical identifier and the imported nodes reconnect without edits.

For the HTTP Request route, store the key under environment variables in the app settings and reference it as shown below, so it never appears inside a node in plain text.

HTTP Request node, as configured in Dify: GET https://api.virlo.ai/v1/content-trends?platform=tiktok&limit=5 | Authorization: Bearer {{#env.VIRLO_API_KEY#}}. Keys come from your Virlo dashboard.

What can a Dify agent research with Virlo?

Paste these into an Agent node instruction or a chat with your Agent app. Replace the brackets with your own details.

  • Chatflow research desk

    When a teammate names a topic, use Virlo to return the top 5 TikTok and Instagram videos on [topic] this month ranked by Virality Score, each with its hook and link.

  • Structured trend JSON

    Return this week's trends to copy in [your niche] as JSON with name, what to keep, what to swap and one example URL, so the next node can parse it.

  • Sound shortlist

    Pick three breakout TikTok sounds and three trending Instagram Reels sounds that suit a [tone] video for [your brand], and say why each fits.

  • Creator vetting

    Look up @[creator] on TikTok with Virlo. How often do their videos beat their usual views, and which three did best?

  • Hook bank

    Collect 15 spoken hooks from outlier videos about [topic], group them by hook type and rewrite each for [your offer].

  • Competitor watch

    Compare @[competitor] and @[competitor 2] on TikTok over the last 30 days: posting pace, best video, and the format each one repeats.

  • Script from a transcript

    Get the transcript of [TikTok link] and rewrite it as a 30-second script for [your product], keeping the beat order.

  • Client recap

    Summarise what changed on TikTok and Instagram in [client niche] this month in five bullets, each with a source video, for a client email.

Dify vs Flowise vs n8n: which guide matches the tool you use?

All three are visual canvases that can call Virlo, but they are built for different jobs.

Dify (this page)

An LLM app platform with Agent apps, Chatflows and Workflows, plus built-in knowledge bases. Virlo is added as a workspace MCP provider.

Flowise

A node-based builder for agents and chatflows on LangChain-style components. Virlo goes into a Custom MCP tool on the Agent node.

n8n

A general automation platform with an AI Agent node. Virlo plugs into its MCP Client Tool, and n8n then moves the result across hundreds of apps.

Dify MCP publishing

Dify can also expose your own app as an MCP server for other clients. That is the reverse direction and isn't needed to use Virlo.

What are the limits?

  • Virlo reads public TikTok and Instagram data only. A Dify agent can research and summarise with it, but it cannot publish, comment or log in to an account.
  • The MCP provider is a workspace-wide setting; changing its server identifier after apps use it breaks those apps, so leave it as virlo.
  • Trends to copy refresh hourly and are checked by the Virlo team; sounds and outlier rankings update through the day.
  • Long research jobs, such as creating a new marketing agent, finish in the background, so ask the Dify agent to check back rather than wait inside one run.
  • For what each Virlo plan includes, see pricing.

Frequently asked questions

How do I add an MCP server to Dify?

Open Tools in the Dify console, switch to the MCP tab and add an MCP server over HTTP. For Virlo, paste https://dev.virlo.ai/api/mcp/mcp, name it Virlo, set the server identifier to virlo, then authorize through the sign-in page or with a bearer header. Its tools then appear in Agent nodes and Agent apps.

Does Dify support MCP?

Yes. Dify connects to remote MCP servers that use HTTP transport and lists their tools as a provider in your workspace. It handles OAuth automatically through Dynamic Client Registration, or you can send custom headers such as an API key. Virlo works through either option.

Can a Dify AI agent see TikTok data?

Not out of the box. Once Virlo is added as an MCP provider, a Dify agent can pull TikTok trends to copy, outlier videos, trending and breakout sounds, hashtags, creator lookups and full transcripts, and use that data inside any chatflow or workflow step.

Can Dify research Instagram Reels?

Yes, through Virlo. The agent can rank Reels on a topic by Virality Score, list the sounds rising on Instagram Reels, read hooks and on-screen text, and check a creator's recent videos. Transcripts are a TikTok feature; Reels return hooks, formats and numbers.

Should I use MCP tools or the HTTP Request node in Dify?

Use MCP tools when an Agent node should decide what to look up from a plain-English instruction. Use the HTTP Request node when a workflow needs the same endpoint and fields every run, such as a Monday pull of content trends into a database.

Does the Dify MCP setup work on self-hosted Dify?

Yes. The MCP tab behaves the same on Dify Cloud and on a self-hosted install, as long as the server can reach dev.virlo.ai over HTTPS. If your instance has no browser for the sign-in step, use the bearer header route with a key.

Where do I get a Virlo API key for Dify?

Sign in to the Virlo developer dashboard, open API Keys and create one. Keys start with virlo_tkn_. You only need it for the header route or the HTTP Request node; the sign-in route lets Dify fetch its own token. Give each Dify workspace its own key so you can revoke one without touching the rest.

Can a Dify workflow post the results to TikTok?

No. Virlo is read-only research and never connects to your social accounts. A Dify workflow can send the findings to Slack, email, a database or any webhook, and you publish with the tools you already use. Keeping research and publishing in separate tools also means a misfiring agent can never post on your behalf.

Is this the same as the Flowise setup?

The server is identical, but the screens differ. In Flowise, Virlo goes into a Custom MCP tool attached to the Agent node with a JSON config, and the key lives in a Flowise variable rather than a workspace provider. The Flowise guide walks through each field and the refresh step that lists Virlo's actions.

Give your Dify agents a social media brain

Add Virlo in the MCP tab once and every app in your workspace can research TikTok and Instagram.