Three Data API Families Built for AI Agents
Most "AI agent" data stacks still boil down to three questions: what does the web say, what is the short-video world doing, and what is happening on X. The three services below each answer one of those questions, and β more interestingly for anyone running Claude Code, Cursor, or an MCP-capable agent β each one ships an agent-native interface alongside the REST API: a /SKILL.md manifest and a hosted streamable-HTTP MCP server.
They are separate products from the same vendor group, priced the same way: pay-as-you-go, no monthly subscription, credits that do not expire, and a small free balance on signup without a credit card.
| Category | Service | What it returns | Base price |
|---|---|---|---|
| Search | SearchData API (searchdataapi.com) |
Google SERP, Trends, News, Autocomplete | $0.20 / 1,000 requests ($0.08β$0.20 on plans) |
| Short-form video | ShortVid API (shortvidapi.com) |
TikTok, Instagram Reels, Meta Ad Library | $0.20 / 1,000 requests (per item on some endpoints) |
| Social (X) | CheapXAPI (cheapxapi.com) |
X/Twitter posts, profiles, search + account actions | $0.0006 / call (β$0.03 / 1K tweets) |
1. Search data: SearchData API
The problem it solves: you need Google-shaped results inside a program β organic rankings, "People Also Ask", 54-week trend curves, and fresh headlines β without running a browser or scraping HTML yourself.
Endpoints (6, all read-only)
| MCP tool | REST endpoint | Cost | Returns |
|---|---|---|---|
google_search |
GET /v1/search |
2 credits | Organic SERP canonical URLs, snippets, related_keywords, People Also Ask, 195 countries |
google_trends_realtime |
GET /v1/trends/realtime |
2 credits | Surging topics, breakout velocity tiers |
google_trends_interest |
GET /v1/trends/interest |
10 credits | 54-week normalized (0β100) interest timeline + rising/top queries |
google_news_search |
GET /v1/news/search |
2 credits | Publisher-attributed news results with UTC timestamps |
google_news_top |
GET /v1/news/top |
2 credits | Breaking headlines by country and topic |
google_autocomplete |
GET /v1/autocomplete |
1 credit | Google Suggest typeahead |
Billing detail worth knowing: empty results and errors are charged 0 credits β you pay for results, not for misses. At 2 credits per SERP call and $1.00 = 10,000 credits, that lands at $0.0002 per search, i.e. $0.20 per 1,000.
The agent interface
- Skill manifest:
https://searchdataapi.com/SKILL.md(MIT licensed) - Remote MCP:
https://mcp.searchdataapi.com/mcp(OAuth 2.1, one-click, orx-api-keyheader) - OpenAPI 3.1:
https://searchdataapi.com/openapi.json
Install into Claude Code in one line:
mkdir -p ~/.claude/skills/searchdataapi && \
curl -fsSL https://searchdataapi.com/SKILL.md -o ~/.claude/skills/searchdataapi/SKILL.md
claude mcp add --transport http searchdataapi https://mcp.searchdataapi.com/mcp \
--header "x-api-key: sr_live_your_api_key_here"
Who this competes with
It positions itself against SerpApi, which starts around $40/month for far fewer searches β the claim is 15xβ75x cheaper per request. The trade-off is scope: SearchData covers SERP, Trends, News and Suggest. It does not try to cover Maps, Shopping, Scholar or Images, so if your agent needs those, this is not a drop-in replacement.
If you are already using JustOneAPI / JustSERPAPI for Chinese-market work, the split is clean: JustSERPAPI and SearchData both do Google SERP for English-language output, while JustOneAPI's core strength is the Chinese platforms (Xiaohongshu, Douyin, Weibo, Taobao, JD). They are complements, not duplicates.
2. Short-form video data: ShortVid API
The problem it solves: programmatic access to TikTok, Instagram Reels and the Meta Ad Library without cookies, logins or residential proxies β including 1080p no-watermark MP4 links.
What you can pull
- Videos β detail (with HD
hdplayMP4), keyword search, regional feeds - Creators β profile, posts, followers, following, creator search by follower count
- Engagement β comments and replies, hashtags/challenges, sounds/music (auto-extracts the audio if you hand it a video URL)
- TikTok Ads β top ads by impression/CTR/like over 7/30/180 days, plus individual creatives
- Instagram β profile, reels, posts (with MP4), comments
- Meta Ad Library β search live Facebook/Instagram ads by advertiser, country, status
That is 21 core endpoints. Unit economics are per item, not per request: videos, reels and creators are $0.20/1K items, comments and followers $0.10/1K, single profile/post lookups $0.001. Failed calls, 429s and empty pages cost nothing.
The agent interface
- Skill manifest:
https://shortvidapi.com/SKILL.md - Remote MCP:
https://mcp.shortvidapi.com/mcp(OAuth 2.1) orhttps://api.shortvidapi.com/mcpwith anx-api-key - OpenAPI 3.1:
https://shortvidapi.com/openapi.json, live latency on/status
Notably, initialize and tools/list respond without authentication, so an agent can discover the tool surface before you provision a key β a small detail that saves a lot of setup friction.
curl -sS "https://api.shortvidapi.com/v1/video/search?keyword=ai+tools®ion=US&count=10" \
-H "x-api-key: $SHORTVIDAPI_KEY"
Honest caveats
This is public-data collection in a category where platforms change shape often. Assume endpoint drift, and validate before you build a product on top of it. The free sandbox (10,500 credits on signup, +20,000 via the dashboard streak) is enough to walk every endpoint once, which is exactly what you should do first.
3. X (Twitter) data: CheapXAPI
The problem it solves: X's official API is priced per tweet read ($5.00 / 1,000), which makes research, monitoring and content pipelines expensive fast. CheapXAPI charges a flat $0.0006 per call, and a search or timeline call returns up to 20 tweets β roughly $0.03 per 1,000 tweets, about 99% below official pricing.
Endpoints
Reads: user info (single and batch by ID), last tweets, advanced search with native X operators, tweets by ID, replies, followers, followings, user search β plus lists, communities, trends, quotes and retweeters. Account actions (posting, liking, following, DMs) are available too, billed the same, with a one-time $0.10 charge for logging the account in.
curl -G "https://api.cheapxapi.com/twitter/tweet/advanced_search" \
--data-urlencode 'query=from:NASA since:2026-01-01 -filter:replies' \
--data-urlencode 'queryType=Latest' \
-H "x-api-key: $CHEAPXAPI_KEY"
The agent interface
- Skill manifest:
https://cheapxapi.com/SKILL.md(MIT) - MCP page:
https://cheapxapi.com/mcp - LLM-friendly reference:
https://cheapxapi.com/llms-full.txt
Caveats that matter
- Free credit is $1.00, valid 60 days β smaller and shorter-lived than the other two.
- Card payments are "coming soon"; top-ups are crypto-only today.
- Account actions mean you are posting as a logged-in account. Treat that like any automation on a platform with terms of service, and keep credentials in environment variables, never in code.
- Rate limits: 2 requests/sec on a new account, 10/sec after the first top-up.
429s are not charged.
How the three fit together
Each service publishes a SKILL.md in the format Claude Code, OpenClaw and other agent runtimes read, plus a hosted MCP server. That means the same three steps work everywhere:
- Drop the
SKILL.mdinto your skills directory β the agent learns the endpoints, auth header and pricing model in one read. - Add the MCP server (OAuth one-click, or an API key header for zero-install).
- Let the agent call tools directly instead of you writing glue code.
For a research or content agent, a sensible combination is:
- Discovery β SearchData for SERP, PAA and trends to find what people ask
- Evidence β ShortVid for TikTok/Reels engagement and ad creatives
- Signal β CheapXAPI for what practitioners say on X, and for publishing
Total cost for a modest workload stays in single-digit dollars per month, because everything is pay-as-you-go and unused credits do not expire.
Before you commit
- Test the free tier first. Each signup grants $1.00β$1.20 plus a sandbox bonus; that covers a full endpoint tour.
- Check data provenance and terms. These are collection-based APIs; confirm they fit your jurisdiction and your product's terms.
- Do not hardcode keys. Use environment variables, and keep them out of commits and skill manifests.
- Read the billing rules. Zero charges on empty results is genuinely useful, but per-item endpoints (videos, comments, followers) bill per record, not per request.
Disclosure: this article is based on the vendors' published SKILL.md manifests and live endpoint checks performed on 2026-11-18; no paid evaluation run has been completed at the time of writing. Pricing and endpoint lists can change β verify against the vendor's own /pricing and /docs pages before relying on them.
