MCP server · mcp.videosupport.io

Give your support agent eyes.

Plug your own AI agent into VideoSupport through one MCP endpoint. It can ask a customer for a video mid-chat, read the transcript and visual triage, then answer with your tools, your rules and your voice.

Get started free See how it works
  • Set up in about 5 minutes
  • Works with Claude, Cursor, VS Code and any MCP client
  • Recordings your agent answers don't use AI reviews
Ticket HS-42 VIA MCP
  1. “My router keeps blinking orange…”
  2. request_recording(“HS-42”, “Show the router lights”)
  3. Customer recorded · 0:24
  4. get_visual_triage(video_id)
  5. Router · 0.94 Likely issue: Network setup
  6. check_warrantyyour tool
  7. reply_to_customer(“HS-42”, …)
  8. Reply sent · human on standby

How it works

From “can you show me?” to answered, in four steps.

VideoSupport handles capture, detection and the handoff to your team. Your agent handles the thinking, using the same tools and docs it already uses.

  1. 01

    Your agent asks for a video

    Mid-conversation, it calls request_recording and sends the customer a link. No app, no install.

  2. 02

    The customer records

    Camera or screen, on any device. Live guidance helps them capture what support actually needs.

  3. 03

    VideoSupport triages

    We transcribe the audio and detect the devices on screen and the likely issue, then hand it to your agent as structured data.

  4. 04

    Your agent answers or escalates

    It runs your business logic, then replies or hands off to a human with the video, transcript and its notes attached.

Your agent → VideoSupport

One MCP server. Every video tool your agent needs.

Six tools, grouped by what your agent is doing. Every call is logged on the ticket.

1 Ask

request_recording(ticket_id, prompt)

Ask the customer to record. Returns a video_id and the recording_url to send them.

2 Watch

get_recording(video_id)

Status, playback and download links, duration, and whether the transcript and triage are ready.

get_transcript(video_id)

Everything the customer said, as full text plus timed segments.

get_visual_triage(video_id)

Devices on screen, the likely issue, possible causes and a summary, before your agent writes a word.

3 Act

reply_to_customer(ticket_id, message)

Answer on the ticket. Replies land in your helpdesk, where your video tickets already live.

escalate_to_human(ticket_id, reason)

Hand off to your team with every video, transcript and triage on the ticket attached.

Get started

Connect your agent in five steps.

Create a workspace, plug the key into your agent, and run a sample conversation to see it work end to end.

  1. Create your workspace

    Your workspace holds your tickets, videos and API keys. Tell us how customers will record.

    How will customers record?

    In VideoSupport, open API in the sidebar and click Create a New Set Of Credentials.

    No VideoSupport account yet? Start free, then come back.

  2. Save your API key

    Your agent uses this key to connect. We only show it once, so copy it or download it now.

    Create your workspace in step 1 to get your key.

  3. Connect your agent

    Pick your client and add the config below. After step 1, your key is filled in automatically.

  4. Send recordings to your workspace

    So your agent can read new videos, VideoSupport needs to notify us when a customer records. Add this webhook URL in VideoSupport.

    Your webhook URL appears here after step 1.
    1. In VideoSupport, open your API settings (or your public link's settings).
    2. Set this URL as the webhook for the API token or public link you used in step 1.
    3. Save. New recordings, transcripts and triage now flow to your agent.

    The URL contains a secret token. Treat it like a password.

  5. Try it

    Check the connection, then run a sample conversation. We play the customer, so you can watch each tool call your agent would make.

    Create your workspace in step 1 to run the test.

    The sample creates a ticket named demo-… in your workspace with made-up video data.

    FAQ

    Questions, answered.

    Which agents can connect?

    Any MCP-capable agent or client that supports remote (Streamable HTTP) servers: Claude, Cursor, VS Code, the OpenAI Agents SDK, LangChain, or your own code. Send your API key as Authorization: Bearer <key>.

    Do I need a VideoSupport account?

    Yes. Recordings are captured by VideoSupport, so you need either an API token or a public link from your account. Agent access through MCP is included on the Pro and Enterprise plans.

    Does this use my AI reviews?

    No. Recordings your own agent answers through MCP don't count against your monthly AI review allowance.

    How does my agent know when the video is ready?

    Call get_recording. Its status changes from requested to recorded, and transcript_status and visual_triage_status become ready when that data arrives.

    Where do replies and escalations go?

    Every reply and escalation is saved on the ticket. With a helpdesk bridge set up, they're also delivered to HubSpot, Intercom, Zendesk, Crisp or Monday.com, signed so your system can verify them. Talk to us to set that up.

    I lost my API key. What now?

    Keys are stored hashed, so we can't show one again. Contact us and we'll issue a new key and revoke the old one.

    Is my data isolated?

    Yes. Each API key only sees its own workspace's tickets and videos, and every action your agent takes is logged on the ticket.

    Your agents. Your rules. Now with video.

    Bring the agent you've already tuned on your docs and systems. We'll handle the video.

    Get your API key