Help centre

Register, verify your email, then create a key in the dashboard. Verification is what releases the free credit, so if your balance looks empty, check that you clicked the link in the signup email — including your spam folder.

Create an API key

Up to 100 per account, which is enough to give every application or project its own. Keys are credentials: keep them secret, and note that deleting one is irreversible.

Send your key as a bearer token in the Authorization header: Authorization: Bearer YOUR_API_KEY. Nothing else is required for a standard endpoint.

First request walkthrough

New accounts receive $0.05 in free credit automatically once the email is verified — about 50 requests at the $0.001 base price. Not every endpoint accepts free credit; paid-only endpoints draw from your paid balance even while free credit remains.

Check endpoint eligibility

No, that was removed. The free credit on registration replaced it.

No. You do not need to supply credentials, cookies, sessions or any other login information from your own social accounts. A small number of endpoints take an extra input such as a user-supplied session token, and where that is the case it is stated in that endpoint's own documentation. If the documentation says nothing, your TikHub API key is all you need.

The Swagger UI at api.tikhub.io is the authoritative, always-current list of every endpoint and its parameters. The written documentation at docs.tikhub.io covers the same ground with more prose.

Open the Swagger UI

TikTok, Douyin, Rednote, Instagram, Twitter/X, YouTube, Threads, LinkedIn, Reddit, Bilibili, Weibo, Lemon8, Zhihu, WeChat Official Accounts, WeChat Channels, Kuaishou, Pipixia and Xigua Video. Several platforms have more than one series — TikTok alone has Web, App V3, Creator, Shop and Ads.

Facebook is not supported because of data regulations. Snapchat and Pinterest are not supported either. If a platform is not in the list above and not in the Swagger UI, we do not offer it.

Yes, fully — tweet detail, user profiles, user tweets, media, replies, comments, retweeters, search, trending, followers and following. For endpoint-level issues, Discord is the fastest route.

Twitter/X API

Yes, through the Web V2 series, but coverage is deliberately narrow: eight endpoints covering user profile and posts, company profile and posts, post detail and comments, job detail, and job search.

LinkedIn API

No. TikHub is data retrieval only. There are no interaction endpoints on any platform, no automated posting, and we do not handle logging in to social platforms on your behalf.

No. There are no monitoring, tracking or surveillance endpoints, and no live-room monitoring. If you need monitoring, build it on top of the retrieval endpoints — the scheduling and alerting is yours to run.

No. We do not disclose how endpoints deal with account risk control, crawler control or anti-crawler measures.

You are. TikHub is a data-access tooling platform: we deliver publicly available social media data through API endpoints. What happens to it downstream — how it is processed, stored, redistributed or built into a product — is outside our control and is your responsibility. That includes complying with each source platform's terms, with data-protection, privacy and intellectual-property law in your jurisdiction, and with any contracts you have with third parties.

Only successful requests. A request is billed when it returns status 200; failed requests are not charged. Prices run from $0.001 to $0.01 per request and differ by endpoint, with the more complex endpoints costing more.

See the price of every endpoint

All pricing is shown in USD. You can pay in other currencies — CNY, EUR, GBP and others — through the supported payment methods.

No. Every request returns fresh, real-time data and is billed independently, even with identical parameters. The cache_url in each response is a separate debugging aid, not a cheaper way to re-fetch.

No. Eligibility is per endpoint, and the API Marketplace and the dashboard pricing table both show which endpoints accept free credit and which participate in tiered discounts.

Check an endpoint

By daily request count, at the $0.001 base price: 0–1,000 is $0.001; 1,000–5,000 is $0.0009 (10% off); 5,000–10,000 is $0.0008 (20% off); 10,000–20,000 is $0.0007 (30% off); 20,000–30,000 is $0.0006 (40% off); 30,000+ is $0.0005 (50% off). The same percentages apply to endpoints with a different base price.

No — they are tiered, the way income tax brackets work. Each bracket of requests is billed at that bracket's rate. At the $0.001 base price, 20,000 requests in one day comes to $15.60: 1,000 at $0.001, then 4,000 at $0.0009, then 5,000 at $0.0008, then 10,000 at $0.0007. It is not 20,000 x $0.0007 = $14.00.

Estimate your cost

Daily, and they apply automatically — there is nothing to enable. The day boundary is Los Angeles midnight, the same clock the usage log runs on.

RPS is how many requests per second your account may send. Every account starts at Level 1, 10 RPS, at no cost. RPS is billed monthly and is entirely separate from what you spend on requests — buying it does not change your per-request price.

Open the RPS dashboard

Level 2 is 20 RPS at $5/month, then 30 RPS at $10, 40 at $20, 50 at $30, 60 at $35, 70 at $40, 80 at $45, 90 at $50, and Level 10 is 100 RPS at $55/month. Above 100 RPS is an enterprise conversation.

You are over your rate limit. Check your current level under API Settings → RPS Settings; plans are monthly and can lapse into an Inactive state, which drops you back to the default. Upgrade or renew to raise the ceiling.

Yes, and it is ours rather than a third party's: 2.6% + $0.30 on Alipay, and 5% + $0.30 on PayPal and USDT. Topping up $5 through PayPal is charged as $5.55. Enterprise accounts are exempt.

Alipay, PayPal, USDT on TRC20 only, and B2B bank transfer for company-to-company payments. Coupon codes, auto top-up and balance alerts all live on the Add Credit page.

Add credit

If you are paying with Alipay from inside mainland China and the payment QR code does not render, connect to a VPN and try again.

Do not log in or sign up when the PayPal page opens. Choose the debit card / credit card option and use guest checkout.

Exact to the last decimal. Send the precise amount shown above the QR code on the payment page, regardless of what your exchange charges — and if your gateway takes a gas fee, add it on top. If the QR code shows $10.5678 and your gateway charges $1.00 in gas, send exactly $11.5678 USDT.

Your own bank's international wire fee, typically around $50, plus a $20 processing fee on our side. B2B transfer is set up through the enterprise flow.

B2B payment

Active codes are shown on the banner at the top of the dashboard Overview page. If there is no banner, there is no active promotion. Enter the code in the field provided and click Apply before completing the purchase.

Dashboard overview

Yes. Invoices for completed orders download straight from Order History, and you can type your own company name or billing title on the download page, so an ordinary company header needs no help from us. We issue a US invoice only.

Download an invoice

That is available if you are an Enterprise customer or a single purchase went over $500 — for example combining several orders, naming a specific billing entity, or adjusting the format. Contact support with your registered email address and the order IDs you want reflected, and we will confirm the details.

Contact support

Only in specific cases, and the policy is strict. A top-up or a dataset order is not refundable by default — a request has to fall inside one of the cases the refund policy sets out. Read it before funding a large balance, and use balance alerts and auto top-up day to day so you are adding what you need rather than a lump sum.

Read the refund policy

Two ways: a single top-up of $3,000, or $4,500 in cumulative top-ups. Either one qualifies — you do not need both.

See enterprise plans

Exemption from the 5% payment processing fee; 3–10% bonus credit on each top-up, scaled by amount and by time since the last recharge; priority custom API development at no extra cost; dedicated technical support in a private Discord channel separate from the public one; and free private deployment tuned for high concurrency. The $3,000 single top-up also carries a $100 bonus.

Not self-serve. Upgrading goes through email or Chatport contact, or by submitting a ticket on the Enterprise page in your dashboard.

Enterprise in the dashboard

If your total request volume will pass 4.5 million a month, contact us separately — that is a custom solution and custom pricing conversation rather than a tier.

Book a call

Yes — $2, redeemed automatically. The catch worth knowing up front: both sides have to be paid users. If either the referrer or the invited friend is still on free credit, the award does not pay out.

Get your referral link

Loyalty points accrue from API spend, and crossing a milestone threshold unlocks cash-credit rewards. Both the points and the thresholds are under Settings → Award Milestones.

Every successful response carries a cache_url that reopens that exact response for 24 hours, free of charge. It is there so you can recover a response your code mishandled, or hand it to a teammate. After 24 hours the link expires and stops working.

No. Accessing the cached result incurs no charge, which is why it is worth saving the cache_url if you expect to read the same data more than once inside the day.

That one needs a human. Raise it on Discord and an agent will look at the specific request — do not keep re-fetching the endpoint in the hope it resolves itself.

Discord

The request format or parameters are wrong — but it is also sometimes a transient server-side error, so always retry before reporting it. When you do report one, include the request URL, the parameters and the response body.

The API token is missing, invalid, unverifiable, inactive, expired, or the user was not found. Check the key is live under API Settings → API Keys.

Insufficient balance. The endpoint requires payment, and depending on the endpoint it may not accept free credit at all.

Missing permissions for the route, a disabled account, an unverified email, or a key that is not authorised for the endpoint you are calling. Check the key's permissions in the dashboard.

API keys

The endpoint URL does not exist in our API. It is not a node issue, not IP blocking, and not a service outage. Verify the path in the Swagger UI; if it is not listed there, we do not offer it.

Check the endpoint exists

A required parameter or field is missing, or a value you sent is the wrong type or format. The endpoint's documentation lists everything it requires.

You are sending requests faster than your RPS level allows. Slow down or raise the level.

A server-side error on our end — the request could not be completed. Retry; if it persists, report it with the request details.

That is normal pagination, not a bug. Paginated endpoints frequently return fewer items than the requested count; keep paginating with the returned cursor.

Thirty or fewer. A larger count on a comment endpoint can cause errors or silently dropped items.

It varies by platform. Douyin search returns roughly 6–10 results per call and the number moves each time. Rednote paginated endpoints return roughly 6–20 items, and the first page may return fewer. WeChat search does not paginate at all — send the same request again for more results, and expect some repeats.

Only ever for an account you own or are authorised to access — never a cookie taken from someone else's account. Open the platform in a browser and log in, press F12 and switch to the Network tab, refresh, pick any request, find the Cookie field in its request headers, and copy the whole value into your API request.

The account-level analytics series: TikTok Creator API and Douyin Creator V2 both operate on one specific creator account and need that account's own valid login cookie. Without it, that account's data cannot be retrieved.

The responses are large and vary per platform, so we document endpoints and parameters rather than every key in the payload. For a specific field, pasting a sample response into Claude or another AI assistant is faster than anything we could write.

Prefer App V3 for stability. Use the Web series only as a fallback when App V3 returns errors.

Not from the user info endpoints — region is not in them. There is a dedicated endpoint, fetch_user_country_by_username, that returns it.

The dedicated Shop series only (/api/v1/tiktok/shop/web/). Shop endpoints that still appear under App V3 or the Web series are legacy, unmaintained, and may return incomplete or stale data without notice. Note that the TikTok Web Shop series and the TikTok Web series are different things despite the similar names.

The TikTok Shop Web API is sometimes unstable or returns intermittent errors. It is usually temporary and normally restored within two to three hours — retry after a short wait before reporting it.

Eight: US, GB, SG, MY, PH, TH, VN and ID. Products from any other region are not supported by that endpoint.

Fetch a TikTok video that contains a product link, then read $.data.anchors[0].extra in the response — that JSON carries both. Note that seller_id is not the same thing as user_id.

TikTok links come in three forms — play, no-watermark and watermark — and the no-watermark content in downloadAddr is the one to prefer. App URLs download directly without cookies; Web URLs need a valid tt_chain_token cookie and return 403 without it. Links also expire in about 24 hours, so a 403 often just means re-fetch the video. Some videos are regionally restricted and need a proxy.

Download walkthrough

fetch_live_room_info only returns data while the room is actually broadcasting; if the stream ended or the room is offline it returns empty or null. Check status first with check_live_room_online. Anonymous viewers and viewers in invisible mode cannot be fetched at all.

Creator account analytics — account health and violations, insight overviews, live and video analytics, video lists, and shop or product performance. They are POST endpoints and each one needs the creator account's own login cookie, so you can only use them for an account you own or are authorised to access.

TikTok Creative Center and Commercial Content data: ad detail and search, top ads spotlight, recommended ads, plus trending products, hashtags, sounds and creative insights. Around 30 endpoints in total, all listed in the Swagger UI.

First verify the video or user is still live and public in the app. Content that was deleted or made private will 400.

App V3 first, Web as the fallback when App V3 errors — with one documented exception, the follower count, below.

Douyin search has its own dedicated series; the search endpoints inside the Web and App V3 series are not functional and should not be used. The dedicated series costs $0.01 per request and carries no volume discount, because of what it costs us to keep running.

Not in the standard video response. Play count comes from a dedicated endpoint, fetch_video_statistics. Note that the figure excludes views bought through Dou+ paid promotion.

Use the Web profile endpoint, handler_user_profile — the App V3 profile does not reliably return the follower count. This is the one exception to preferring App V3. The follower list is likewise the Web endpoint.

Retry with the Web version of the video fetch, then check $.data.filter_list[0].reason. Code 8 means the content is unavailable on copyright grounds overseas (short dramas, film clips) or was deleted; 5 means private; 10 means partially visible to author-selected users only. For other codes, contact support.

Anything under /api/v1/douyin/creator_v2. Those are account-level analytics for a creator's own account and require that account's valid Creator-Platform cookie in the POST body. Without access to the account, the data cannot be retrieved.

V1 is public Creator Service Center data and needs no login — activities and campaigns, the material-centre billboards (hot videos, trends, topics, props, challenges, music), creator courses, topic-related videos, user search, mission lists and per-video danmaku. V2 is private, cookie-authenticated analytics for the creator's own account: per-post stats, play-source breakdown, search-keyword stats, watch trend, comment analysis, audience portraits, their own posted-item list, live broadcast history, and an AI-style account diagnosis.

Xingtu (星图) is Douyin's official influencer-marketing platform. The series returns KOL marketing analytics: base info, audience and fans portraits, service price, data overview, video performance, conversion analysis and the Xingtu index, plus keyword KOL search. The typical flow is to resolve the creator's Xingtu kolid from a uid, sec_user_id or unique_id, then call the kol_* analytics endpoints. A newer V2 series adds ranking lists, the creator marketplace, business-card info and similar-creator recommendations.

Douyin's 榜单 rankings — hot lists by category, city and rising popularity, the activity calendar, audience portraits and comment word clouds, hot accounts with fan trends and portraits, and video, topic and search leaderboards including low-fan breakouts and high completion, like and follow rates.

抖音指数, modelled on ByteDance's 巨量算数. Around 44 endpoints for keyword, creator, video, brand, topic and content-creation trend analytics: hot-topic rankings and hot words, multi-keyword heat trends, related-word analysis, crowd portraits, creator comparison and similar-creator discovery, brand index and radar charts, and packaged recommendation and search-trend insight reports.

No. Douyin has no shop or product API. For e-commerce data, TikTok Shop is the only option, and it is TikTok-only.

Live room data beyond the standard endpoints is the Live Room Code product. Anonymous viewers and viewers in invisible mode cannot be fetched.

Live Room Code

App V2, with Web V3 as the fallback for web-style note and search access. The old App and Web series were fully retired on 13 June 2026 and their endpoints are disabled; Web V2 is likewise no longer maintained.

Yes, and it is worth the five minutes — Rednote is the most quirk-heavy series we run, and the guide covers correct usage and the platform's own oddities before you hit them.

Rednote usage guide

Easiest: you do not need one. App V2 endpoints accept a share_text parameter instead of note_id or user_id, and take xiaohongshu.com, xhslink.com and xhslink.cn links. Otherwise, open the note or profile on the website (not the app) and read the ID out of the URL path — after /discovery/item/ for a note, after /user/profile/ for a user. Both are 24-character hex strings.

From that same web URL's query string — xsec tokens are no longer returned in API responses. URL-decode it first if it ends in %3D, which is an encoded trailing equals sign. Web V3's note detail endpoint needs both note_id and xsec_token; it has no share_text option, so if all you have is a share link, use App V2 instead.

That is a bug on the Rednote platform itself, not something we can fix. Relatedly, get_video_note_detail returns three notes in its response — the one matching the note ID or URL you passed in is the one to use.

No, and not because of us: Rednote does not expose those metrics at all. Note content, images, videos, user profiles, comments, likes, shares and search results all work normally.

$0.01 per request, with no volume discount unless you are doing a million or more requests a month. Notes are fetched one at a time — there is no batch note endpoint.

No. Topic posts and notes-by-tag are not supported, and the old get_notes_by_topic endpoint is unmaintained. Use keyword search instead.

Retry several times first. If it still fails, backend resources are loading — wait 30 to 60 minutes before trying again.

The V2 series, across three namespaces: wechat_search/v2 for search, wechat_mp/v2 for Official Accounts, and wechat_channels/v2 for Channels. The old Official Accounts and Channels series were fully retired on 13 June 2026 and rebuilt.

All of it lives in one series, with two endpoints. fetch_search is the general 搜一搜 search covering official accounts, articles and Channels videos in a single call, with business_type selecting the vertical (all, account, article, video, sticker). fetch_search_videos is for Channels videos with filters: duration, sort order and publish time. The per-series search endpoints that used to exist are all retired.

It splits cleanly. The eight article-level endpoints take the article's web URL — no ID lookup needed. The three account-level endpoints take the account's gh_ username.

fetch_article_detail_h5 is the recommended default: more reliable body text and the most complete field set, including title, author, cover, publish time and 合集 info. fetch_article_detail is faster with fewer fields. They are independent endpoints — neither falls back to the other.

Set a 30-second client timeout. WeChat's own servers are slow and the H5 endpoints fetch and parse the article page; too short a timeout can bill you without returning a response.

comment_id, mid and similar values exceed JavaScript's safe integer range of 2^53-1. Read and pass them as strings — never through Number.

fetch_video_detail returns the media address plus a decode_key. Supply at least one of object_id, export_id or share_url — priority runs in that order. You do not need an object_id to start: pass the share link and the response returns the object_id along with everything else. Set raw=false for the simplified structure, which is the one to use for downloads.

Two separate things. Anti-hotlinking means the bare url often will not open — concatenate url + url_token, or use the ready-made full_url. And if the file downloads but will not play, it is encrypted: decrypt it with the decode_key from the same response. WeChat issues a new link and key on every request, even for the same video, so always use a matching pair from one response.

Decryption tool

With raw=false, $.data.media is a single object holding url, url_token, full_url and decode_key. With raw=true the shape is $.data.objects[0].objectDesc.media[0], camelCased: url, urlToken and decodeKey.

No. WeChat does not expose those metrics publicly.

Because an Instagram post accumulates engagement across both Instagram and Facebook, and we return only the Instagram side. What the app shows is the combined total. Our number being smaller is expected, not an error.

V2 is the sensible default — the most stable and detailed of the visitor-view series. But V1 and V2 are both visitor-view, and their figures, especially counts and statistics, can lag reality by a few days. When you need real-time data, use V3.

If the discrepancy is not explained by the Instagram-versus-Facebook split above, contact support with the specific post and what you expected.

Contact support

Use the V2 stream endpoint, get_video_streams_v2. That is the endpoint built to return download and stream info.

Download URLs returned by the other YouTube endpoints can only be accessed, viewed and downloaded from a California residential IP. From anywhere else you need a California residential proxy — or use get_video_streams_v2 instead, which is the recommended route.

Some YouTube endpoints take need_format. Set it to true and you get cleaned, structured data that is much easier to parse.

The Bilibili video API does not return high-quality video URLs. That is a platform limitation, not something we can work around.

Bilibili uses DASH, so video and audio arrive as separate .m4s files that you download separately and merge with FFmpeg. Requests need a valid Referer header or they return 403. URLs are time-limited to roughly two hours. HD and 4K need login cookies such as SESSDATA, and 1080p and above require a premium membership.

Download guide

Both follow the same pattern — App API first, then the Web series as a fallback. Weibo's fallback is Web V2.

Eight endpoints: user profile and posts, company profile and posts, post detail and comments, job detail, and job search. The older LinkedIn Web series is fully retired.

Most sub-resource endpoints: per-section profile data (skills, education, experience, certifications, honours, publications, recommendations, volunteering, images, videos, bio, top card, contact info, follower and connection counts, followed groups and companies, recent activity); company extras (employees and count ranges, grouped locations, similar companies, competitors, stock quote, CTA buttons, job counts and company job lists); post extras (reactions, reposts, comment replies, post-by-slug, hashtag feed); discovery endpoints; group info and posts; the Ad Library; and every search endpoint except job search. If you need one of those, it is not currently available — and there is no substitute endpoint to suggest.

No. The sora2/create_video endpoint in the social media API is deprecated and unmaintained. Sora video generation now lives in AI Proxy.

AI Proxy

A marketplace of pre-processed, ready-to-use public social media datasets. The data is cleaned and updated regularly, so you buy and download rather than scrape. The page has two tabs: Marketplace to browse and buy, My Datasets for your orders and downloads. Be clear on one thing before ordering: most of this data was collected at random rather than to a schedule, and it is not real-time. If you need current data for a specific account, post or query, that is the API, not a dataset.

Explore the API Marketplace

Pick a dataset and preview it — up to 300 sample rows, with the full set delivered after purchase. Each listing shows its data fields, total records, price and minimum order, and a View Pricing estimator gives the total for a given record count. Choose an output format and volume, then pay in full.

Browse the catalogue

No. The order is processed and delivered automatically to the My Datasets tab. When the status reaches Completed, a Download button appears there.

My Datasets

Twenty-four hours. The link expires after that and the file is removed, so download promptly once the order completes.

Every order with its order ID, task ID, dataset, format, amount, status, progress percentage, row count, ETA, any error, the date, and a Download or Details action. Status covers both Completed and Failed.

CSV, JSON, JSONL and Parquet, chosen at purchase.

Around 23 datasets across more than ten platforms. The larger ones: TikTok video 161M+, user 15M+, comment 12M+, hashtag 14M+ and product 2M+; Douyin video 94M+, user 12M+ and comment 487M+; Rednote note 42M+; X tweet 11M+ and comment 27M+; Instagram post 2.5M+ and comment 3M+; Bilibili video 9M+; YouTube video 3M+; plus WeChat Channels and Kuaishou video sets. Those figures are a snapshot taken in October 2026 — every set keeps growing, and new sets are added, so each listing shows its own live total.

Keyword search in the Search datasets box, plus filters by platform (TikTok, Douyin, Bilibili, Instagram, Rednote, WeChat, Kuaishou, X, YouTube, Weibo) and by type (video, user, comment, hashtag, live, shop, product, post, note, tweet, channel, status).

Yes, with Advanced filters on the preview page. Keyword search matches rows containing any of up to five phrases, and field filters let you write up to ten WHERE rules against the dataset's own columns — which vary per dataset. You get ten filtered preview queries per dataset.

Volume-tiered and charged per 1,000 records, with a $150 minimum order. The per-record rate falls as the record count rises.

For the TikTok Video dataset: 100K–500K at $2.00 per 1K; 500K–1.33M at $1.60; 1.33M–5M at $1.20; 5M–20M at $0.80; and 20M+ at $0.40. Tiers vary by dataset, so each listing's own cost estimator is the number to trust.

Production-ready Python source code for connecting to TikTok and Douyin live-room WebSockets. It connects to a room, parses the binary message stream, and delivers real-time danmaku and interaction events. It is source code you run yourself, not a hosted endpoint.

Live Room Code

Basic interaction — viewer enter, likes, chat danmaku, room status, plus emoji and common-text messages on Douyin. Engagement — follows and fan reward points, with fan-club interaction on Douyin. Monetisation — gifts, live shopping, co-host sync and co-host fan rewards, PK and team-PK battles. Data services — room statistics, hourly rank and room ranking.

Fifteen paid WebSocket message APIs for TikTok across four groups, and twenty for Douyin plus an extra Activity Operations group.

WebSocket connection logic, heartbeat and auto-reconnection, Protobuf binary parsing and Gzip decompression, plus one year of technical support.

Pricing is per-API — buy the individual message types you need or a full package — and there is no self-serve checkout. Contact us, or book a call directly and pick a slot.

Book a call

A Python environment. The product is the connection and parsing layer; you integrate it into your own analytics or dashboard tooling.

It connects AI agents — Claude, Cursor, VS Code Copilot, Cline, Cherry Studio, anything MCP-compatible — straight to the data tools, with over 990 tools across 16 platforms served by dedicated per-platform servers.

Set up MCP

Douyin 247, TikTok 204, Instagram 82, Rednote 71, Weibo 64, an Others server (Lemon8, PiPiXia, Xigua, Toutiao, Sora2) 64, Bilibili 41, YouTube 37, Kuaishou 33, Zhihu 32, LinkedIn 25, Reddit 24, TikHub utilities 23, WeChat 19, Twitter 13, Threads 11.

Stdio for Claude Desktop, VS Code Copilot Chat and Claude Code — it bridges HTTP to stdio through npx mcp-remote and needs Node.js. Streamable HTTP for Cherry Studio, Cursor, Cline, curl and custom clients; it is direct HTTP and the most efficient. SSE for legacy clients that need a streaming connection.

One server per platform. Streamable HTTP is https://mcp.tikhub.io/{platform}/mcp and SSE is https://mcp.tikhub.io/{platform}/sse, both authenticated with your API key as a bearer header. Stdio bridges to the /mcp URL. There are also two unauthenticated meta-endpoints, /health and /platforms.

Add one mcpServers block per platform, with command npx and args mcp-remote, the platform's /mcp URL, then --header and your bearer token. Claude Desktop's config file is at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS and %APPDATA%\Claude\claude_desktop_config.json on Windows. The MCP page has a configurator that generates the exact block. Restart the client after editing.

Yes, for a single platform at a time, with the tikhub-mcp CLI — run it with a platform, the sse transport and a port, then point your client at http://localhost:8001/sse.

Yes, via langchain-mcp-adapters. Install langchain-mcp-adapters and langgraph, then register each platform with transport streamable_http, the platform's /mcp URL, and the bearer header.

No — add only the ones you use. Each platform is a separate server, and loading all of them makes the agent's tool list unwieldy.

No, and this catches people out. AI Proxy runs on a separate account and a separate billing system. Credits and balances are not shared with the social media API in either direction — you register and top up separately.

AI Proxy

One gateway for OpenAI, Google Gemini, Anthropic Claude, DeepSeek, Sora, Seedance, Kling and Veo, spanning text, reasoning, multimodal, image, embedding, audio and video generation. Seven providers in total.

13% to 71% below official rates depending on the provider: OpenAI and Sora around 71% off, Seedance and Kling 65%, Claude 57%, Gemini 46%, Veo 35%, DeepSeek 13%.

No. AI Proxy is not compatible with the official SDKs — not OpenAI's, not Anthropic's, not Google's. Use direct HTTP requests: curl, fetch, axios, requests. Set the base URL to https://ai.tikhub.io, authenticate with your key, and keep the same request and response shapes as the official APIs.

Most endpoints take Authorization: Bearer YOUR_API_KEY. Two exceptions follow their provider's convention: the native Claude Messages API uses x-api-key, and Gemini uses x-goog-api-key.

Sign up at ai.tikhub.io, check model availability and pricing on its pricing page, take your key from the console token page, and try requests interactively on the developer portal.

AI Proxy docs

It is currently in beta. For issues or to request an additional model, contact support.

A public showcase of interactive applications built on the APIs: TikTok video metrics, fake-view detection, comment analysis, creator analysis, a voice-cloning tool, and an open-source desktop downloader for TikTok, Douyin, Instagram and YouTube.

Open the demo showcase

They are free to open, but they run on your own API key and spend your credits. They are demonstrations of what the API can do, not separate free services. The voice-cloning demo also needs an ElevenLabs key of your own.

No. The Chrome extension, the MCP servers, Zapier and n8n all call the same API. Usage through any of them consumes requests and draws down the same balance, exactly as a direct call would.

Zapier connects the APIs to 5,000+ apps for no-code, trigger-based automation. n8n is the self-hosted, open-source alternative, with full control of your data and custom workflows. Both are available now.

Integrations

Account home: total balance, free credit, today's usage in API calls, spend and endpoints used, trend and spending charts, and your top endpoints. Active promotion banners appear here too.

Overview

Browse and test over 1,000 endpoints across 16+ platforms, filtered by platform, category, API or WEB type, free-credit eligibility and discount eligibility. It shows each endpoint's cost and rate limit — it is the place to check before you assume an endpoint takes free credit.

API Marketplace

The Usage Log has per-day history with request counts and a cost breakdown, a date-range filter, and CSV or JSON export.

Usage log

The usage log runs on Los Angeles time (America/Los_Angeles). Daily totals roll over at midnight there, so a request can land on a different calendar day than it does in your own timezone — and daily volume discounts reset on that same clock.

Yes, under Price & Billing. It estimates daily cost across endpoints and applies the progressive volume discounts from 0% to 50%. Estimates only, but they use the same tier logic as the biller.

Cost calculator

Yes. There is an endpoint for balance, credits and account status, and another for today's usage. Both are in the TikHub utilities series.

User endpoints

The free demo and analytics tools, the paid TikTok and Douyin Live WebSocket source code, and the enterprise contact link. Items marked Paid are paid.

Usually Cloudflare's web verification not having run. Refresh the page to trigger it, then log in again.

The email is not verified yet. Click the link in the signup email — and check spam if it is not in your inbox. Verification is also what releases the free credit.

Your browser has probably cached an old dashboard URL. Click Return to Home to go back to the homepage and navigate in again.

Only Enterprise accounts can. Regular accounts need a mainstream provider such as Gmail, Outlook or Yahoo. A company-domain account has to be created by us by hand, and if you already have a personal account we can move its remaining balance across.

Forgotten password goes through the forgot-password page and sends a reset link to your registered email. To change a known password, use Settings → Account → Change Password, which asks for the current one.

Forgot password

Promotional email, yes, under Email Preferences. Account-essential email — verification, password reset, payment — is always sent.

Check it under API Settings → API Keys: an expired or inactive key causes authentication failures. Also confirm the key is authorised for the endpoint you are calling. Never expose a key, including the bearer token embedded in a generated MCP config.

Settings → Account → Delete Account schedules permanent deletion with a seven-day cancellation window. After that window it cannot be undone.

No. The TikHub API balance and the AI Proxy balance are separate and are not shared. Top-ups on either side fall under the same refund policy, so fund each one for what you will actually use.

Read the refund policy

Add Credit has both a balance alert and auto top-up. Setting one of them is considerably less painful than discovering a 402 halfway through a batch.

Add credit

Several channels, each with its own response time. The contact page lists every one of them, with the hours they are staffed — check there for the details rather than guessing from here.

All the ways to reach us

Product updates and announcements go out on X, and we send a monthly email notice to registered accounts.

@TikHubio

The getting-started guide: how to get your API key and make a first request, end to end.

Getting started

docs.tikhub.io is the complete API reference, and the Swagger UI at api.tikhub.io is the interactive explorer with the authoritative endpoint list.

API documentation

Step-by-step guides live on the tutorials page, and the blog carries longer walkthroughs — including the video download guide referenced throughout these answers.

Tutorials

Yes, on GitHub: the Python SDK, the multi-functional desktop downloader, demo projects and more.

GitHub