Early access launch offerNew accounts get 200 free credits. Purchases get double credits.

Let your app and agent actually watch social media content and pull out the details

mamitsu is an API and MCP server for Instagram, TikTok, YouTube and X posts. Send a link and what you want to know. mamitsu watches the video, listens to the audio and returns the answer as JSON.

@miso.udonInstagram
15-minute one-pot creamy miso udon
Ready
instagram.com/p/DN4qQLfgSR1
Prompt

Get the recipe from this post: the dish, the ingredients, the toppings and the steps in order.

{
  "dish": "15-Minute One-Pot Creamy Miso Udon",
  "time_minutes": 15,
  "ingredients": [
    "miso", "peanut butter", "udon noodles",
    "egg", "spinach"
  ],
  "toppings": [
    "sliced green onions", "black sesame seeds",
    "crispy chilli oil"
  ],
  "steps": [
    "Whisk miso and peanut butter together in a pot.",
    "Add udon noodles to the broth.",
    "Poach an egg directly in the broth.",
    "Add a handful of spinach to the pot."
  ]
}
InstagramTikTokYouTubeXClaude CodeCodexHermesPi

Most of a post never makes it into the caption

The small, important details are said out loud or flash on screen for a second: the quantities, the street name, the price. mamitsu watches the whole post, so you can ask for them in plain words. Add a JSON Schema when your code needs the same fields every time.

Get the recipe from this post: the dish, the ingredients, the toppings and the steps in order.

{
"dish": "15-Minute One-Pot Creamy Miso Udon",
"time_minutes": 15,
"ingredients": [
"miso", "peanut butter", "udon noodles",
"egg", "spinach"
],
"toppings": [
"sliced green onions", "black sesame seeds",
"crispy chilli oil"
],
"steps": [
"Whisk miso and peanut butter together in a pot.",
"Add udon noodles to the broth.",
"Poach an egg directly in the broth.",
"Add a handful of spinach to the pot."
]
}

Try it out yourself

REST API

Turn social media posts into useful data for your users

  1. 1Someone posts a reelA creator shares 13 things to do in Osaka
  2. 2Your app sends one requestThe reel URL and a prompt, in one API call to mamitsu
  3. 3Your users get a mapEvery place comes back as data your app can use
Requestcurl
curl https://usemamitsu.com/api/v1/extract \
  -H "Authorization: Bearer $MAMITSU_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://www.instagram.com/p/Cxbh8I8Ie26",
    "prompt": "List every place in this reel and its kind."
  }'
instagram.com/p/Cxbh8I8Ie26
POST /api/v1/extract
{
  "trip": "13 Things to do in Osaka, Japan",
  "places": [
    { "name": "Dotonbori", "kind": "neighborhood" },
    { "name": "Kuromon Ichiba Market", "kind": "food" },
    { "name": "Osaka Castle", "kind": "castle" },
    { "name": "Shinsekai", "kind": "neighborhood" },
    { "name": "Tsutenkaku Observation Deck", "kind": "viewpoint" },
    { "name": "Namba Yasaka Jinja", "kind": "temple" },
    …
  ]
}
MCP server

Paste a reel into your agent and ask about the video

Add the mamitsu MCP server to Claude Code, Codex, Hermes, Pi or any MCP client and sign in once. When a link comes up, your agent calls mamitsu and gets back only the fields it asked for, so a ten-minute video costs a few lines of context instead of a full transcript.

  • "Go through these creator videos and tell me which ones mention our brand and what they say"
  • "Save the restaurants from these TikToks to my Notion"
  • "What gear does this YouTube review recommend, and at what price?"
Add mamitsu to your agent Browse agent plugins on GitHub

Why I built mamitsu

I kept saving reels of places, recipes and absolutely anything, and never opened them again. What I wanted was inside the video. I first built mamitsu as part of a more niche app, then decided to release it on its own. Now my AI agents keep my recipe library and my trip ideas from the reels I send them, and I've built a couple of apps on the API too.

Unai, maker of mamitsu

FAQ

Can an AI agent watch a TikTok or Instagram video?

Yes, with mamitsu. Connect the MCP server and your agent can send a video link to the extract tool. mamitsu watches the video, listens to the audio and reads the caption, and your agent gets the answer back as JSON.

How is mamitsu different from a video transcription API?

A transcription API gives you the words spoken in the video. mamitsu also looks at the frames and reads the caption, and returns the specific fields you ask for, like the products on screen or the steps of a recipe.

Can I get the same JSON fields every time?

Yes. Send a JSON Schema with the request and mamitsu returns an object in that shape. Without one, mamitsu picks descriptive fields from your prompt.

Can mamitsu summarize a social media video from its URL?

Yes. Ask for a summary in your prompt, with the points you care about, and mamitsu returns it as JSON. You send the link and mamitsu fetches the video for you.

How much does an extraction cost?

Each extraction uses credits based on the length of the video or the number of photos. A 45-second reel uses 1 credit. New accounts start with 200 free credits. During the early access launch offer, purchases get 2 credits per cent. The pricing page has more examples.

Start with that post you saved last week