Arctic Shift API and PullPush API: endpoints, parameters, limits.

Two free APIs replaced Pushshift. Arctic Shift has the richer one: search, ID lookup, comment trees, aggregations and time series, fresh to within seconds, with keyword search limited to one user, subreddit or thread at a time. PullPush keeps the old Pushshift query style and can search comments across all subreddits, but it is often unavailable and its data was reported to stop in May 2025. Neither gives an uptime guarantee. This page lists what each endpoint takes, what the errors mean, and how an old Pushshift query maps onto both, and onto a commercial archive when free is no longer enough.

Are the Arctic Shift and PullPush APIs up?

On our last check (12 November 2026 at 11:07 UTC), the Arctic Shift API was responding (HTTP 200 in 1.8 s). The newest post it returned was 18 seconds old. On our last check (12 November 2026 at 11:07 UTC), the PullPush API was not responding (HTTP 526).

Both are checked every hour on the status page.

Arctic Shift API endpoints

The base URL is arctic-shift.photon-reddit.com and no key is needed. Source: the project's API documentation on GitHub, read on 10 November 2026. Check it before you build: the author marks parts of it as subject to change.

Arctic Shift search parameters and their limits

• Keyword parameters need a scope. On posts, title, selftext and query require an author or a subreddit. On comments, body requires an author, a subreddit, a link_id or a parent_id. Without one, the API answers HTTP 400 with the message: 'body' query parameter requires one of: author, subreddit, link_id, parent_id (recorded on 10 November 2026).
• Very active users and subreddits. The documentation says keyword search is not supported on them, and that aggregations on them have a high chance of timing out.
• limit. From 1 to 100 on search, 25 by default. The value auto returns between 100 and 1,000 results depending on server capacity.
• Dates. after and before accept epoch seconds or milliseconds, ISO 8601, or an offset from now such as 2d or 1year.
• fields. A comma separated list, to return only the fields you need.
• Fresh items. score and num_comments read 0 or 1 at first and are updated after roughly 36 hours.

Two errors account for most failed calls. "Query timed out" means the parameter combination is too heavy: reduce the limit or add a filter, and try once more, since the documentation notes the database sometimes needs to warm up. HTTP 429 means you are rate limited: the X-RateLimit-Reset and X-RateLimit-Reset-At headers give the wait. Limits are dynamic, based on server load and query complexity.

PullPush API endpoints

PullPush keeps the Pushshift shape: a comment search and a submission search on api.pullpush.io, under /reddit/search/comment/ and /reddit/search/submission/. The familiar parameters apply: q for the keyword, author, subreddit, after, before, size and sort, with ids or link_id for direct lookups. Old Pushshift scripts usually port by changing the base URL.

The limits are operational more than technical. The rate limit is roughly 30 requests per minute, outages are frequent, in August 2026 the API was answering automated queries with HTTP 429, and the ihsoyct project noted in October 2025 that PullPush data stopped at 20 May 2025. The details are on our PullPush page.

Mapping an old Pushshift query

The THINKPOL column lists capabilities only. Its endpoint names, parameters and commercial terms are not public: the documentation is shared with vetted accounts.

Porting code: what actually changes

1. The base URL. For PullPush that is often the only change.
2. Parameter names, on Arctic Shift. q becomes query, title or selftext on posts and body on comments. size becomes limit. Results are always sorted by creation date.
3. The scope rule. Every keyword query to Arctic Shift needs an author, a subreddit or a thread. A sitewide keyword sweep has to be redesigned or sent elsewhere.
4. Pagination. Page by date: take the creation time of the last item and pass it as the next after or before.
5. Failure handling. Back off on HTTP 429, retry a timeout once, checkpoint long jobs.
6. Provenance. Store the source and the retrieval time with every result. "Returned nothing" is only meaningful next to the name of the archive that returned it.

Which one should you build on?

For a study with a fixed window, neither API: take the Arctic Shift dump files, which do not move. For lookups on a known user or subreddit, the Arctic Shift API. For legacy Pushshift code that only needs to keep running on older data, PullPush, with retries.

For anything that has to work after March 2027, slow down. Reddit closes public access to its Data API that month, and Arctic Shift states that it collects through the official Reddit API. We found no statement about the closure in its documents on 10 November 2026, and PullPush's site sat behind a challenge page for our check that day, so we could not read what it has published. Do not put a deadline on a dependency whose future is not written down.

When free is no longer enough

The free APIs are the right answer until one of three things becomes unacceptable: a keyword that cannot be searched across Reddit, a gap you cannot measure, or a service nobody answers for. THINKPOL covers those three: 30 billion posts and comments since 2005, content deleted by authors and removed by moderators retained, full-text search across the whole archive in under 300 ms, a live firehose, on a supported API that is also exposed over MCP for agents.

It is commercial and metered, and access is vetted and contractual. You can test the data first, with no account, through archive search, user lookup and Reddit trends. Then request access.

How this page is made

The Arctic Shift endpoints, parameters, limits and error messages are taken from the project's API documentation on GitHub, read on 10 November 2026, and from requests we ran that day. The PullPush parameters are the Pushshift-style ones used by the public tools that still query it; its limits are those we have measured and published since July 2026. The status lines are refreshed every hour by an automated check. Third-party APIs change without notice: treat this page as a map and the projects' own documentation as the reference. Corrections: contact us.

Reddit archive APIs, answered.

Does Arctic Shift have an API?

Yes. It is free, needs no key, and covers search, ID lookup, comment trees, aggregations, subreddit and user data and time series. On our last check (12 November 2026 at 11:07 UTC), the Arctic Shift API was responding (HTTP 200 in 1.8 s). The newest post it returned was 18 seconds old.

What is the Arctic Shift API rate limit?

There is no fixed number. Limits are calculated from server load and query complexity. The documentation says a couple of requests per second is fine for a normal user, and an HTTP 429 response includes X-RateLimit-Reset headers giving the wait.

Why does the Arctic Shift API return "Query timed out"?

The parameter combination is too heavy for the server, which usually happens with the keyword parameters body, selftext or title. Reduce the limit, add a date range, a subreddit or an author, and try again once.

Can I search comments by keyword with the Arctic Shift API?

Only inside a scope. The body parameter requires an author, a subreddit, a link_id or a parent_id, and the documentation says it is not supported on very active users or subreddits. A keyword across all of Reddit is not possible with it.

Is the PullPush API still working?

On our last check (12 November 2026 at 11:07 UTC), the PullPush API was not responding (HTTP 526). PullPush is run on community resources, so outages are frequent, and the ihsoyct project noted in October 2025 that its data stopped at 20 May 2025.

Is there a replacement for the Pushshift API?

Not a complete free one. PullPush keeps the same query style with partial data, Arctic Shift offers a richer API with scoped keyword search and monthly dumps, and commercial archives such as THINKPOL provide full-text search across the whole archive with deleted content retained.

THINKPOL is an independent intelligence platform and is not affiliated with, endorsed by, or sponsored by Reddit Inc. or any third-party tool named on this page. Platform and tool names belong to their respective owners. Third-party tool descriptions reflect publicly observable functionality on the verification date stated on the page and may change; corrections are welcome via our contact page.