X API Review (2026): Pricing, Limits and What twitterapi.io Does
An outside read on the X API reseller a lot of developers end up using, built from twitterapi.io's own published numbers and the caveats buyers leave in reviews.
See Synexa models → Official site
What twitterapi.io actually sells
The first line on its homepage is a disclaimer: independent third-party service, not affiliated with X Corp, trademarks belong to X Corp. Everything after that is programmatic access to public social data. The navigation points at docs, a pricing page, a payment page, a QPS limits page, a stream product with its own latency figures, and a page of tweet filter rules. The quickstart is a single curl call against api.twitterapi.io with an X-API-Key header, hitting an advanced search endpoint. There is also a native MCP server, pitched as a way to give live data to Claude, GPT or Cursor without writing glue code. Read it as a data layer rather than as a cheaper official tier and your expectations will land in the right place.
Pricing, in the numbers the site publishes
Signup comes with $0.1 in free credits and no credit card. After that the headline is $0.15 per 1,000 units, described as linear per-call pricing with no monthly floor and no read cap. Real-time data is listed separately, starting at $29 a month. The site's own comparison table puts the official API at $100 or more per 1,000 units and other providers at $0.55 or more, which is a vendor claim about competitors rather than an audited benchmark. One reviewer on Trustpilot mentions an academic arrangement in which half their credits were reimbursed after the paper was published. Rates on a page like this move, so open the pricing and QPS pages before you commit a budget to a number you read here.
Speed claims, and the caveats reviewers add
Throughput is the loudest part of the pitch: 1,000 or more requests per second, median response under half a second, and a stream widget in the header quoting P50 251ms and P90 327ms. The site claims 99.99% uptime and more than 10,000 developers, and points at a 4.6 Trustpilot rating across 28 verified reviews plus roughly 21,000 developers and 94 reviews on its Apify actors. The useful detail sits in the less glowing reviews. One paying user reports latency occasionally spiking to a minute, and a repost, undo-repost, repost-again sequence losing the second event. Another notes that the outages they hit traced back to X itself rather than the service. Neither is disqualifying, but both belong in your retry logic.
Where this is the wrong tool
Anything that needs sanctioned write access, a support agreement with the platform, or compliance sign-off on where the data came from should go through official channels instead, whatever the price difference looks like. A third-party layer can also change shape underneath you: endpoints get renamed, event types get dropped, and you have no contractual claim when they do. The other common mismatch is subtler. Plenty of teams buy social data because they want to feed it into a model, and then discover the expensive, fiddly part was never the data at all. If that is your project, price the inference side separately before you scale the scraping side.
What you actually get
Free credits to test
The $0.1 in starting credits with no card attached is enough to run a batch of searches and see whether the response shape fits your parser before anyone signs off on spend.
MCP without glue code
A native MCP server means an LLM client can query live data directly. If you were planning to write a wrapper service for exactly that, this removes a small project.
Linear per-call billing
No monthly floor and no read cap means a quiet month costs almost nothing. It also means a runaway loop costs real money, so cap your own concurrency.
A third party, stated plainly
The homepage says it is not affiliated with X Corp before it says anything else. That honesty is welcome, and it is also the main risk you are accepting.
Two different purchases, side by side
| Feature | twitterapi.io | Synexa |
|---|---|---|
| What comes back | Public posts, user data and a live stream | Model output: images, video and audio |
| Billing shape | Per call, from $0.15 per 1,000 units as listed | Per run, no subscription to hold open |
| Interface | REST plus a native MCP server | One REST endpoint plus a Python SDK |
| Getting started | $0.1 in free credits, no card required | Create a key and call the endpoint |
| Relationship to the platform | States plainly it is not affiliated with X Corp | Runs the hosted models it serves |
| Reach for it when | You need social data at volume | You need FLUX, video or audio generation behind one API |
Getting started in four steps
- Claim the free credits
Sign up for a key and spend the $0.1 in starting credits on real queries rather than the sample one. Malformed edge cases tell you more than the happy path. - Run one search end to end
Point the advanced search endpoint at a query you actually care about, then parse the response into your own schema. Field names are where most migrations quietly break. - Read the QPS page before scaling
Concurrency limits live on their own page for a reason. Check them before you write a worker pool that assumes the 1,000 requests per second headline applies to your key. - Model the bill at your real volume
Multiply your expected calls by the per-1,000 rate on the pricing page, then add the real-time tier if you need the stream. Do this before the trial credits run out.
FAQ
Is this the official X API?
No. twitterapi.io states on its own homepage that it is an independent third-party service and not affiliated with X Corp, and that the Twitter and X marks belong to X Corp. Anything that requires a sanctioned relationship with the platform, such as write access or compliance attestation, has to go through official channels regardless of what a reseller charges.
How much does the X API cost through twitterapi.io?
The site advertises $0.15 per 1,000 units as its starting per-call rate, with no monthly minimum and no read cap, plus $0.1 in free credits at signup. Real-time streaming data is listed separately from $29 a month. Those are the figures published at the time of the crawl, so confirm them on the pricing page before budgeting.
How reliable is it in practice?
The site claims 99.99% uptime and sub-second median responses. Public reviews are largely positive but add texture: one buyer reports occasional latency spikes to about a minute, another reports a repost-undo-repost sequence dropping an event, and a third says the interruptions they saw tracked X's own outages. Build retries and idempotency accordingly.
Can I plug it straight into an LLM?
Yes, that is one of the stated selling points. A native MCP server is offered so that clients like Claude, GPT-based tools or Cursor can pull live data without a custom integration layer. If your architecture already speaks MCP, this removes a wrapper service you would otherwise have to build and maintain.
What if I need to generate media, not collect posts?
That is a separate purchase and worth separating in your budget. A social data API prices per record retrieved; a model API prices per generation run. Synexa is the latter: one REST endpoint and a Python SDK across FLUX, video and audio models, billed per run, so a pipeline that was never really about scraping does not pay scraping prices.
Who should skip this category entirely?
Teams with compliance requirements about data provenance, anyone who needs to post rather than read, and projects small enough that the official free allowances already cover them. Third-party access is a practical answer to a pricing problem, not a legal or contractual equivalent to the real thing, and it is worth being clear about which one you need.
One endpoint for image, video and audio models
If the social data was only ever the input and the hard part is generating something from it, Synexa runs that half: one REST call, a Python SDK, and you pay per run instead of holding a seat.
See Synexa models →