Twitter/X public data, no official developer platform

If you want public tweets as JSON, the official path is the X developer platform: an app, OAuth, review, a SKU. That is the right path if you are building on X. It is a slow path if you need search, a timeline, or a follower list tonight and you are not applying to become a partner.

twitter.fetcher.sh is the other path. You call a GET. Public tweets, handles, search, timelines, and follower lists come back in the usual { status, message, data } envelope. You do not wait in an app-review queue.

x.com is still X. The host is just the API. Private accounts and DMs are not in the catalog, and this is not a firehose contract. It is a priced JSON API for public objects.

What you can fetch

Search, including X’s own operators when you put them in query. Tweets by id. User timelines, replies, retweeters, lists, trends. Resolve a handle to a profile. Pull followers and followings. Every path is GET. Prices and parameters are on the host and in /docs.

A first call, with a prepaid key:

curl -H 'Authorization: Bearer bby_live_…' 'https://twitter.fetcher.sh/api/search?query=from:OpenAI'

Without a key the same URL answers 402. Pay per call and how an agent pays cover the settlement. from:, since:, until: is the operator deep dive.

Most list endpoints want a numeric id. Resolving a handle is the lookup. Follower graphs are the edges. Brand mentions reuse search. None of those jobs require Elevated access.

How this feels compared to the developer portal

The portal is an identity and a contract. You are an app. You have scopes. You might post. You might pay for a firehose. That is a product company relationship.

This host is a URL. You are a client of fetcher.sh. You read what is already public. You do not post. You do not read DMs. You do not get a compliance archive. If a tweet is gone, the envelope says so in JSON instead of a spinner on x.com.

Teams stall because they think they must apply in case they need write access later. Apply when you need write access. Fetch when you need a tweet. Mixing those decisions is how a research task waits six weeks.

Google’s site:x.com is not a substitute. It is a SERP. Use google.fetcher.sh for the open web; use this host for tweets.

A case: a reporter reconstructing a public thread

A reporter has a tweet URL and a handle. They need the tweet, the replies under it, and a week of that account’s public timeline — not to republish a firehose, but to sit the quotes next to each other.

They resolve @OpenAI to an id. They fetch the tweet by id. They pull replies. They run search with from:OpenAI since:2026-01-01 until:2026-01-08. None of that requires Elevated access. None of it returns DMs or protected accounts. If a tweet is gone, the envelope says so in JSON.

That is a morning’s work with a key from /topup. The official platform is still the right tool if they need write access or a compliance firehose. For public reconstruction, the host is enough.

A second reporter on the same desk only has a search string, not a URL. They start with operators instead of a tweet id: from:OpenAI since:2026-01-01 until:2026-01-08 -filter:replies. They page until the window is covered, store tweet ids, and only then fetch replies on the ones that matter. Search is the net. Tweet-by-id is the close read. Mixing those two in one confused GET is how you get a page of Latest results you already had.

A case: an agent that should not store an X app secret

A small agent searches from:OpenAI when a user asks what the lab posted. The developer does not want an X client secret in the prompt or in a shared repo.

They use @x402/fetch or a bby_live_ key. The agent calls search, maybe resolves a handle, maybe pulls a tweet by id. It never completes OAuth. It never sees a DM. If the user wanted to post, they would need the official platform, and this agent would refuse.

A newsroom weekly that is still not a firehose: one from: query for the accounts on the beat, one keyword query with since: for the last seven days, store tweet ids, fetch replies only on the ones that became the story. Lists and trends exist on the host when the question is a list or a trend. They are not a substitute for search.

DMs, protected accounts, and write access remain official-platform work. Public JSON remains this host. The 402 you see without a key is checkout, not a ban.

When to use the official platform instead

Use developer.x.com if you post as a user, if you need private data, or if you have already signed a firehose contract. We return public objects. A failed call is JSON, not an HTML error page. Docs live on the host and at /docs; agents can read /skill.md.