CrewAI MCP: give your crew live TikTok and Instagram research
Updated
Virlo is social listening for TikTok and Instagram Reels. CrewAI agents can split a job into roles and hand work to each other, but nothing in a default crew knows which TikTok or Instagram videos broke out this week. Add Virlo to an agent's mcps list and that agent can pull trends to copy, outlier videos, hooks and sounds from a database of 23.1M viral videos for the rest of the crew. CrewAI is an open-source Python framework for orchestrating role-playing AI agents that work together as crews and flows.
- One `MCPServerHTTP` entry in an agent's `mcps` list
- Give research tools only to the scout, not every agent
- Sequential crews that go from data to scripts in one kickoff
Part of Virlo's AI assistant guides. Want a single agent or a graph instead of a crew? See LangChain. Building on Anthropic's harness? See the Claude Agent SDK.
How do I add the Virlo MCP server to a CrewAI agent?
Five steps in a Python project. CrewAI connects to Virlo when the crew starts and exposes its research tools only to the agents you list it on.
Step 1: Create your Virlo account for free
Step 2: Install CrewAI and the MCP package
The mcps DSL, which CrewAI recommends for MCP work, needs the
mcppackage next tocrewai. Use uv or pip, whichever your project already uses.Terminaluv add crewai mcpStep 3: Store the key in your .env file
Create a key in the Virlo developer dashboard and add it to the
.envfile at your project root, the file CrewAI projects already load for model keys..envVIRLO_API_KEY=virlo_tkn_YOUR_KEYStep 4: Give the scout agent Virlo through mcps
Import
MCPServerHTTPfromcrewai.mcp, point it at Virlo with your bearer header, and list it in the scout'smcps. Streamable HTTP is the default for this config, so no transport flag is needed. The writer gets no tools and works only from what the scout hands over.crew.pyimport os from crewai import Agent, Task, Crew, Process from crewai.mcp import MCPServerHTTP virlo = MCPServerHTTP( url="https://dev.virlo.ai/api/mcp/mcp", headers={"Authorization": f"Bearer {os.environ['VIRLO_API_KEY']}"}, ) scout = Agent( role="Social trend scout", goal="Find TikTok and Instagram formats worth copying in [your niche]", backstory="Marketer who only trusts ideas backed by real videos", mcps=[virlo], ) writer = Agent( role="Script writer", goal="Turn the scout's findings into short video scripts", backstory="Writes tight 30-second scripts for [your brand]", )Step 5: Kick off the crew
Add a research task for the scout and a script task for the writer, then call
kickoff(). Verbose mode prints each Virlo tool call as the scout makes it.Your first promptresearch = Task( description="Use Virlo to list this week's top 5 trends to copy and 5 outlier videos in [your niche].", expected_output="A ranked list with links, what to keep and what to swap", agent=scout, ) scripts = Task( description="Write one 30-second script per trend from the research.", expected_output="Five scripts, each citing its source video", agent=writer, ) crew = Crew(agents=[scout, writer], tasks=[research, scripts], process=Process.sequential) print(crew.kickoff())
Need only one Virlo tool on an agent? CrewAI's DSL lets you filter a server down to named tools, which keeps a narrow agent focused and its context small.
mcps DSL vs MCPServerAdapter vs custom tool: how should CrewAI reach Virlo?
| CrewAI route | What it looks like | Best for |
|---|---|---|
mcps DSL with MCPServerHTTP | Agent(..., mcps=[MCPServerHTTP(url=..., headers=...)]) | Most crews: connection handled per agent, tools discovered automatically |
MCPServerAdapter (crewai-tools) | A context manager that yields MCP tools you pass to tools= | Older crews already built around crewai-tools and manual connection control |
Custom BaseTool on the REST API | A tool class that calls https://api.virlo.ai/v1/content-trends with a bearer header | A Flow step that must return the same fields every run |
| Virlo CLI from a tool | A tool that shells out to virlo trends digest --json | Quick prototypes on a machine where the CLI is set up |
The DSL is the shortest route and the one CrewAI documents first; the other three are there when a crew needs more control.
Which agent in a CrewAI marketing crew should hold Virlo?
| Agent role | Gets Virlo? | Why |
|---|---|---|
| Trend scout | Yes | Finds trends to copy, outliers and sounds; everything downstream depends on this data |
| Competitor analyst | Yes | Needs creator lookups and outlier videos for named handles |
| Script writer | No | Writes from the scout's output, which keeps scripts tied to sourced ideas |
| Editor or reviewer | Optional | Can re-check a source video's hook and Virality Score before sign-off |
Which TikTok trends would a CrewAI scout report today?
This is the live list a Virlo-equipped scout agent reads when its task asks for this week's trends to copy: the trends with the most creators posting them right now.
In Virlo's tracking of 39 live trends to copy, "My moon my man couple montage" leads TikTok this week with 659 creators posting it as of November 19, 2026.
As of November 19, 2026 at 13:07 UTC
| Trend | Stage | Creators this week | Creators last week |
|---|---|---|---|
| Fresh | 659 | new this week | |
| Rising | 346 | 115 | |
| Fresh | 175 | new this week | |
| Fresh | 158 | 33 | |
| Fresh | 158 | new this week |
See all 39 trends to copy
Recipes, 30-day charts, client fit and every example video are in the Virlo app.
Open Trends to copyHow this sample is made
- Source
- Virlo's Trends to copy: formats found by Virlo's trend detector and published only after the team checks each one.
- Window
- Creators who posted each trend in the last 7 days, counted from TikTok and Instagram videos only. Trends made with an AI tool are left out.
- Method
- Ranked by creators posting this week. Stages: Fresh, Rising, Peaking, Fading.
- As of
- November 19, 2026 at 13:07 UTC. Refreshes every hour.
- Made by
- Numbers are pulled automatically from Virlo's data every hour; the text around them is written by the Virlo team.
Suggested citation: Source: Virlo, “CrewAI MCP: give your crew live TikTok and Instagram research”, virlo.ai/crewai
How do I run a weekly content crew with CrewAI and Virlo?
A four-agent crew covers the whole loop from research to a ready brief. One structure marketers use:
- Scout (with Virlo): pulls this week's trends to copy and breakout sounds for [your niche].
- Analyst (with Virlo): looks up two or three competitors and lists the videos that beat their usual views.
- Writer (no tools): drafts a script per trend in your brand voice, citing the source video.
- Planner (no tools): lays the scripts into a seven-day calendar with a sound for each post.
- Run it from a CrewAI Flow on a schedule, and save the final output to a file or send it to your team chat.
Because Virality Score measures how far a video beat its own creator's average, the scout's picks favour formats that work for small accounts, not just videos that rode a large following.
What changes when a CrewAI crew moves to production?
In development the key sits in .env; in a deployed crew, put VIRLO_API_KEY in the platform's secret store and read it the same way. The MCPServerHTTP line stays as it is.
CrewAI caches the tool list it discovers, so repeated kickoffs skip re-discovery. Restart the process to pick up new Virlo tools after a release.
Give each crew its own Virlo key if several teams share one deployment. Revoking one key in the Virlo dashboard then stops only that crew.
What task descriptions work for a CrewAI agent with Virlo?
Use these as Task descriptions for the agent that holds Virlo. Replace the brackets.
Scout task
Use Virlo to find the 5 trends to copy with the most creators in [your niche] this week. For each, say what to keep, what to swap and give one example video.
Analyst task
Look up @[competitor] on TikTok and Instagram. List their three best videos of the last 30 days by Virality Score and the format each uses.
Sound task
Find three breakout TikTok sounds and three trending Instagram Reels sounds that suit a [tone] video for [your brand].
Hook task
Collect 12 spoken hooks from outlier videos about [topic], tag each with its hook type and pass them to the writer.
Transcript task
Fetch the transcript of [TikTok link] and outline its beats so the writer can rebuild it for [your product].
Creator task
Find 10 TikTok creators in [your niche] under [N] followers whose videos often beat their follower count, for a partnership list.
Hashtag task
Report which hashtags in [your niche] gained the most TikTok videos this week and which are cooling, so the planner can tag each post.
Reviewer task
Check every script's source link with Virlo and flag any idea whose video scored under 18 on Virality Score.
CrewAI vs CrewAI AMP vs LangChain: which guide fits?
The CrewAI name covers an open-source framework and a hosted platform. Here is where Virlo fits in each.
CrewAI framework (this page)
The open-source Python library with Agents, Tasks, Crews and Flows. Virlo is added in code through mcps.
CrewAI AMP
CrewAI's hosted platform for deploying and monitoring crews. A crew deployed there keeps the same MCPServerHTTP line, with the key stored as a secret.
LangChain
A separate framework centred on single agents and LangGraph graphs, with its own MCP adapter.
OpenAI Agents SDK
OpenAI's lightweight agent library, which loads Virlo as a hosted or Streamable HTTP MCP tool.
What are the limits?
- Virlo is a read-only source: your crew can study public TikTok and Instagram videos with it, never publish, comment or log in.
- Only agents that list Virlo in
mcpscan call it; the rest of the crew sees no Virlo tools, which is usually what you want. - Background jobs such as a new marketing agent finish later, so split them into a Flow step that checks back instead of waiting inside one task.
- Trends to copy refresh hourly after a check by the Virlo team; sounds and outliers move through the day.
- For what each Virlo plan includes, see pricing.
Frequently asked questions
How do I add an MCP server to a CrewAI agent?
Pass it in the agent's mcps list. For a remote server, import MCPServerHTTP from crewai.mcp and give it a url and optional headers. For Virlo, use the dev.virlo.ai MCP URL with an Authorization bearer header, and the agent discovers its tools at kickoff.
Does CrewAI support streamable HTTP MCP servers?
Yes. MCPServerHTTP uses Streamable HTTP by default, and the older MCPServerAdapter accepts a streamable-http transport too. Virlo is a hosted Streamable HTTP server, so the default configuration is all it needs: a url and an Authorization header. There is no local process to start, which keeps deployed crews simple.
What are CrewAI tools for social media research?
CrewAI tools are functions an agent can call during a task. Adding Virlo gives an agent research tools for TikTok and Instagram: trends to copy, outlier videos ranked by Virality Score, sounds, hashtags, creator lookups and TikTok transcripts. The tools appear to the agent with names and descriptions, so its role and goal decide when it reaches for each one.
What is the difference between mcps and MCPServerAdapter in CrewAI?
The mcps field is the simple DSL: you declare servers on the agent and CrewAI manages connections and tool discovery. MCPServerAdapter, from crewai-tools, is the advanced route where you open the connection yourself and pass the tools in. New crews usually start with mcps.
Can a CrewAI crew research TikTok trends automatically?
Yes. Put Virlo on a scout agent, give it a task that asks for this week's trends to copy in your niche, and run the crew from a scheduled Flow. Each kickoff returns fresh trends with example videos, and downstream agents such as a writer or planner turn them into scripts and a calendar without extra wiring.
Where do I put the Virlo API key in a CrewAI project?
Add VIRLO_API_KEY to the project's .env file during development and read it with os.environ in the MCPServerHTTP headers. In production, store it in your deployment's secret manager. Keys start with virlo_tkn_ and come from the Virlo developer dashboard. Avoid pasting the key straight into crew.py, because CrewAI project folders are often committed to Git.
Can CrewAI agents post to Instagram through Virlo?
No. Virlo reads public TikTok and Instagram data and never connects to social accounts. A crew can produce scripts, captions and calendars from it, and posting stays with your own tools. That split also makes crews safe to run unattended, since no agent holds a login to your brand accounts.
Is CrewAI or LangChain better for Virlo?
Both connect in a few lines. CrewAI fits role-based teams of agents that pass work along; LangChain fits a single agent or a custom graph. If your plan is a scout handing work to a writer, CrewAI maps onto it directly. The LangChain guide covers its adapter for the other case.
Can a CrewAI Flow react when Virlo finds something new?
Yes. Register a Virlo webhook for events such as tracking.outlier_video.detected or content_research_agent.run.completed and point it at an endpoint that starts your Flow. The crew then kicks off the moment a tracked creator posts an outlier or a marketing agent run finishes, instead of on a fixed timer.
Give your crew a social media brain
One MCPServerHTTP line puts live TikTok and Instagram research in your scout agent.
More AI assistant guides
- AI assistant guides
- Claude
- Claude Code
- Cursor
- MCP server
- TikTok MCP server
- Instagram MCP server
- ChatGPT
- ChatGPT dots
- OpenAI Codex
- Gemini CLI
- Grok Build
- Claude Cowork
- n8n
- Zapier
- Grok Bot
- Muse
- Claude Duet
- Hermes Agent
- Kimi K3
- Cue
- Manus
- Perplexity Computer
- Paperclip
- OpenClaw
- NanoClaw
- DeepSeek
- AI agents CLI
- ChatGPT agent
- Perplexity
- Microsoft Copilot
- Le Chat
- Notion
- Slack
- Discord
- Telegram
- Google Sheets
- Windsurf
- VS Code
- GitHub Copilot
- Cline
- JetBrains
- Kiro
- Google Antigravity
- goose
- Warp
- OpenCode
- Zed
- Amp
- Factory Droid
- Raycast
- Lovable
- Replit
- Bolt.new
- v0
- Ollama
- LM Studio
- Open WebUI
- LibreChat
- Make
- Pipedream
- Gumloop
- Lindy
- Relevance AI
- Airtable
- ClickUp
- Dify
- Flowise
- LangChain
- OpenAI Agents SDK
- Mastra
- Vercel AI SDK
- Claude Agent SDK