Home›Guides›Limits

What this tool cannot do

Every limit and all sixteen failure codes, read from the code that serves this site. Last reviewed 2026-10-28.

This page exists because the useful question is not what an exporter promises, it is what it does when the answer is no. Everything below is read from the source, not estimated, including the parts that make this tool look worse than a competitor that says nothing.

The one thing that will break your client

A failing lookup or export returns HTTP 200. The body carries {"ok": false, "errorCode": "..."}. The only responses that are not 200 are the two rate-limit refusals, one per endpoint. If your code branches on the status code, every other failure will look like a success and you will write a null document to disk. Read ok first.

This is not an accident of the framework — it is how these two endpoints were written. A request that reaches the handler and is then refused still answers 200, because the refusal is a normal outcome of a lookup rather than a fault in the request. The 429 exists separately because a rate limit applies before any work is attempted, and the lookup and export budgets are counted apart from each other.

The 13 limits

Each row names the constant behind it, so you can check rather than trust. If a number here disagrees with what the page does, the page is right and this table is stale — tell us and it gets fixed the same day.

1Post lookups per address — 30 every 5 minutesRATE_LIMIT_PARSE = (30, 300) in core/config.py. Counted per client address, sliding window, in process memory.
2Exports per address — 80 every 5 minutesRATE_LIMIT_EXPORT = (80, 300). Separate budget from lookups — running out of lookups does not block an export you already previewed.
3Posts per thread — 40MAX_THREAD_POSTS = 40. A thread longer than this is exported as its first 40 posts and the truncation is reported, not silently dropped.
4Links in one paste — 12MAX_INPUT_URLS = 12. Each line counts as one link, even if two lines point at the same post.
5Images per document — 60MAX_IMAGES = 60. Images beyond the limit are left out of the file and counted in the warning, not quietly skipped.
6Size of one image — 12 MBMAX_IMAGE_BYTES = 12 * 1024 * 1024 = 12582912. Measured on the bytes actually received, so a wrong Content-Length cannot slip a larger file through.
7Text per document — 400,000 charactersMAX_TEXT_CHARS = 400000. Roughly 60,000 words; nothing on X comes close, so in practice this limit never fires.
8Time to reach X — 20 secondsREQUEST_TIMEOUT = 20. Per attempt, not per lookup — with a proxy pool a single lookup can spend up to three of these.
9Time to fetch one image — 25 secondsIMAGE_TIMEOUT = 25. Images are fetched one at a time, so one slow host delays the export rather than failing it.
10Retries per lookup — 3 exit attemptsATTEMPTS_WHEN_POOLED = 3 in core/proxy.py. Only applies when a proxy pool is configured; with no pool there is exactly one attempt.
11Retryable upstream statuses — 403, 407, 429, 503ROTATE_STATUS = {403, 407, 429, 503}. On these the request is retried through a different exit. Any other status is returned as-is, because another IP would get the same answer.
12How long a document lives — 3 hoursDOC_TTL_SECONDS = 3 * 3600 = 10800. After that the docId no longer resolves and the export answers doc_expired.
13Documents kept at once — 60MAX_STORED_DOCS = 60. This is a cap on the whole server, not per visitor — the oldest document is evicted first, so a busy hour can expire yours early.

All 16 failure codes

The second column is the HTTP status you actually get. Read it carefully: only one of these is not 200.

1invalid_url · HTTP 200 · Yours · retry: NoThe link is not an X post, article or thread URL. Both x.com and twitter.com are accepted, including the m. and mobile. forms. A link to a profile, a search page or a hashtag is refused here rather than fetched.
2article_id_link · HTTP 200 · Yours · retry: NoThe link is an /i/article/<id> form. Those ids are internal to X and are not addressable through any public read endpoint, so nothing can be fetched from them. Paste the /handle/article/<id> form instead — that one has a real post id behind it.
3not_found · HTTP 200 · X's · retry: NoX answered 404. The post is deleted, the handle was renamed, or the id was mistyped by one digit. Nothing to retry.
4protected · HTTP 200 · X's · retry: NoX answered 401 or 403. The account is private, the post is age-gated, or the post is limited to approved followers. This tool never signs in to X, so a protected post is a final answer, not a bug waiting to be fixed.
5rate_limited · HTTP 429 · X's, or ours · retry: LaterOne of the only two failures that come back as a real 429 — the lookup endpoint and the export endpoint each have their own counter, so both can produce this code. Three quite different things land here: your lookup budget running out, your export budget running out, or X throttling our outbound address. The body does not distinguish them, so waiting is the only useful response.
6timeout · HTTP 200 · Ours · retry: YesA request to X did not finish within the 20 second timeout, on every attempt. Worth trying again — the usual cause is a slow exit, not a broken post.
7upstream · HTTP 200 · Ours · retry: YesX answered something we do not have a specific code for. Retrying often works, because the next attempt may leave through a different exit.
8http_NNN · HTTP 200 · X's · retry: DependsThe literal number X returned, folded into the code, for example <code>http_451</code>. Returned when the status is neither 404, 401, 403 nor 429. Retry only if the number is 5xx; a 4xx will answer the same way forever.
9too_many · HTTP 200 · Yours · retry: NoMore links or more posts than the run accepts — over 12 links in one paste, or a thread longer than 40 posts. Split the input instead of retrying.
10too_large · HTTP 200 · Yours · retry: NoThe document would exceed the text ceiling of 400,000 characters. In practice this only happens when several very long articles are merged in one paste.
11no_content · HTTP 200 · X's, or yours · retry: NoThe post exists but yielded nothing exportable. Usually a post that is only an image, only a video, or only a poll. Also what you get when every image was unticked before exporting.
12nothing_selected · HTTP 200 · Yours · retry: NoThe export was submitted with every image deselected. Re-select at least one, or export as Markdown, which does not embed images.
13doc_expired · HTTP 200 · Ours · retry: NoThe docId is older than 3 hours, or it was evicted because the server was already holding 60 documents. The fix is to look the post up again — the docId cannot be revived.
14bad_format · HTTP 200 · Yours · retry: NoThe requested output format is not one of docx, pdf, md, txt or zip. A typo in the format field lands here.
15missing_output · HTTP 200 · Ours · retry: YesThe export ran and reported success, but the file was gone by the time it was read back. This is a server-side race, not something you did. Retrying is correct.
16server · HTTP 200 · Ours · retry: YesAn unhandled exception. Deliberately kept as a distinct code rather than being reported as a bad link, so a wrong link is never blamed for our bug.
Retrying, in one sentence: retry timeout, upstream, missing_output, server and any http_5xx. Do not retry anything else — a wrong link, a private post and an expired document all answer identically on the tenth attempt.

What it deliberately does not do

1It does not download video or images as media filesThe output is a document. A video inside a post becomes a link in that document, not an MP4. Images are embedded in DOCX and PDF, and written to a folder in the ZIP export — but there is no path that hands you a video file.
2It does not sign in to X, and cannotEvery fetch is anonymous. That is why protected posts are a final answer, and why nothing here can see a post that is limited to followers.
3It does not rewrite, summarise or translateThe text goes into the document as written. If the post is in Japanese, the document is in Japanese. No model is involved anywhere in the pipeline.

What is stored, and for how long

1Uploaded links and generated documentsKept for 3 hours, then deleted. At most 60 documents exist at any moment, across all visitors, oldest evicted first.
2Request logsKept 30 days. Records the event, locale, format and a request id. Not the document text.
3AccountsThere are none. No sign-up, no cookie identifying you, no profile of what you exported.

Questions this page gets asked

Does a failing export return an HTTP error code

Almost never. The only non-200 responses are the two rate-limit refusals — one on the lookup endpoint, one on the export endpoint. Every other failure is returned as HTTP 200 with a body of {"ok": false, "errorCode": "..."}. If you are writing a client, read ok and errorCode — checking the status code alone will make every other failure look like a success.

Why is article_id_link a separate error instead of a bad URL

Because the link is well formed and looks valid. An /i/article/<id> link carries an internal X id that no public endpoint accepts, so it cannot be fetched at all. Reporting it as invalid_url would send people hunting for a typo that is not there.

Can it export a protected or private post

No. The fetch is anonymous; there is no login and no cookie. A protected post returns the protected code immediately rather than being retried, because retrying cannot change the answer.

What happens to a thread longer than 40 posts

The first 40 posts are exported and the truncation is reported in the warning field. Nothing is silently dropped, and the document does not pretend to be complete.

Is the output really editable, or is it a screenshot

Really editable. DOCX gets real heading styles and real paragraphs, so Word's navigation pane works and text can be reflowed. Nothing is rendered to an image anywhere in the pipeline.

How long do I have to download the file

Three hours from the moment the document was created, and less if the server is busy — at most 60 documents are kept at once across all visitors and the oldest is evicted first. After that the docId answers doc_expired and the post has to be looked up again.

Does it work for posts in languages other than English

Yes. The interface is in English, Chinese, Spanish and Portuguese, but the exported text is whatever language the post was written in. Nothing is translated.

Why did my images come out missing

pbs.twimg.com sometimes refuses the request outright. When that happens the image is reported as missing in the warning field rather than replaced with a placeholder, so you can tell a broken document from a complete one.

Is there an API

The same two endpoints the page uses, and nothing more: a lookup and an export. No key, no account. They are subject to the per-address budgets above, and they are not versioned — treat them as internal.

Can I merge several posts into one document

Yes, up to 12 links, one per line. They appear in the order you pasted them, not in chronological order, so put them where you want them in the file.

Does it keep my documents

For three hours, then no. Documents are files on the server, not rows in a database, and they are deleted on the same timer whether or not anyone downloads them.

Checking any of this yourself

Look the same post up twice. The first call costs one unit of the 30-per-5-minutes budget; a repeat inside the cache window does not. Send a deliberately wrong link and watch the status line: it will say 200, and the body will say invalid_url. That single test tells you whether a client is reading this API the way it was written.

Choose the next check after an incomplete result

Compare the last visible post in the preview with the reply you intended to keep. A missing source post and a file that fails to open require different checks. To test your reader independently, open the site-generated Word example. For a missing reply, use the preview-to-file comparison.