Skip to content
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Go to file
This makes it more clear that the `pageSize` param controls the number of zipnum blocks examined and *not* the number of actual results/CDX lines per page.

I also tried to clean up the language in general in the pagination section to read a little better.
4 contributors

Users who have contributed to this file

@ikreymer @Mr0grog @nemobis @aponb

Wayback CDX Server API - BETA

Table of Contents

Intro and Usage

Advanced Usage

Intro and Usage

The wayback-cdx-server is a standalone HTTP servlet that serves the index that the wayback machine uses to lookup captures.

The index format is known as 'cdx' and contains various fields representing the capture, usually sorted by url and date.

The server responds to GET queries and returns either the plain text CDX data, or optionally a JSON array of the CDX.

The CDX server is deployed as part of Wayback Machine and the usage below reference this deployment.

However, the cdx server is freely available with the rest of the open-source wayback machine software in this repository.

Further documentation will focus on configuration and deployment in other environments.

Please contant us at for additional questions.

Basic Usage

The most simple query and the only required param for the CDX server is the url param

The above query will return a portion of the index, one per row, for each 'capture' of the url "" that is available in the archive.

The columns of each line are the fields of the cdx. At this time, the following cdx fields are publicly available:


It is possible to customize the Field Order as well.

The the url= value should be url encoded if the url itself contains a query.

All other params are optional and are explained below.

For doing large/bulk queries, the use of the Pagination API is recommended.

Url Match Scope

The default behavior is to return matches for an exact url. However, the cdx server can also return results matching a certain prefix, a certain host or all subdomains by using the matchType= param.

For example, if given the url: and:

The matchType may also be set implicitly by using wildcard '*' at end or beginning of the url:

  • If url is ends in '/*', eg* the query is equivalent to
  • if url starts with '*.', eg url=* the query is equivalent to

(Note: The domain mode is only available if the CDX is in SURT-order format.)

Output Format (JSON)

 ["org,archive)/", "19970126045828", "", "text/html", "200", "Q4YULN754FHV2U6Q5JUT6Q2P57WEWNNY", "1415"],
 ["org,archive)/", "19971011050034", "", "text/html", "200", "XAHDNHZ5P3GSSSNJ3DMEOJF7BMCCPZR3", "1402"],
 ["org,archive)/", "19971211122953", "", "text/html", "200", "XAHDNHZ5P3GSSSNJ3DMEOJF7BMCCPZR3", "1405"]]
  • By default, CDX server returns gzip encoded data for all queries. To turn this off, add the gzip=false param

Field Order

It is possible to customize the fields returned from the cdx server using the fl= param. Simply pass in a comma separated list of fields and only those fields will be returned:



A new form of filtering is the option to 'collapse' results based on a field, or a substring of a field. Collapsing is done on adjacent cdx lines where all captures after the first one that are duplicate are filtered out. This is useful for filtering out captures that are 'too dense' or when looking for unique captures.

To use collapsing, add one or more collapse=field or collapse=field:N where N is the first N characters of field to test.

Query Result Limits

As the CDX server may return millions or billions of record, it is often necessary to set limits on a single query for practical reasons. The CDX server provides several mechanisms, including ability to return the last N as well as first N results.

  • The CDX server config provides a setting for absolute maximum length returned from a single query (currently set to 150000 by default).

  • Set limit= N to return the first N results.

  • Set limit= -N to return the last N results. The query may be slow as it begins reading from the beginning of the search space and skips all but last N results.


  • Advanced Option: fastLatest=true may be set to return some number of latest results for an exact match and is faster than the standard last results search. The number of results is at least 1 so limit=-1 implies this setting. The number of results may be greater >1 when a secondary index format (such as ZipNum) is used, but is not guaranteed to return any more than 1 result. Combining this setting with limit= will ensure that no more than N last results.

    Ex: This query will result in upto 5 of the latest (by date) query results:

  • The offset= M param can be used in conjunction with limit to 'skip' the first M records. This allows for a simple way to scroll through the results.

    However, the offset/limit model does not scale well to large querties since the CDX server must read and skip through the number of results specified by offset, so the CDX server begins reading at the beginning every time.

Advanced Usage

The following features are for more specific/advanced usage of the CDX server.

Resumption Key

There is also a new method that allows for the CDX server to specify 'resumption key' that can be used to continue the query from the previous end. This allows breaking up a large query into smaller queries more efficiently. This can be achieved by using showResumeKey= and resumeKey= params

  • To show the resumption key add showResumeKey=true param. When set, the resume key will be printed only if the query has more results that have not be printed due to limit= (or max query limit) number of results reached.

  • After the end of the query, the will be printed on a seperate line or seperate JSON query.

  • Plain text example:

org,archive)/ 19970126045828 text/html 200 Q4YULN754FHV2U6Q5JUT6Q2P57WEWNNY 1415
org,archive)/ 19971011050034 text/html 200 XAHDNHZ5P3GSSSNJ3DMEOJF7BMCCPZR3 1402
org,archive)/ 19971211122953 text/html 200 XAHDNHZ5P3GSSSNJ3DMEOJF7BMCCPZR3 1405
org,archive)/ 19971211122953 text/html 200 XAHDNHZ5P3GSSSNJ3DMEOJF7BMCCPZR3 1405
org,archive)/ 19980109140106 text/html 200 XAHDNHZ5P3GSSSNJ3DMEOJF7BMCCPZR3 1402
 ["org,archive)/", "19970126045828", "", "text/html", "200", "Q4YULN754FHV2U6Q5JUT6Q2P57WEWNNY", "1415"],
 ["org,archive)/", "19971011050034", "", "text/html", "200", "XAHDNHZ5P3GSSSNJ3DMEOJF7BMCCPZR3", "1402"],
 ["org,archive)/", "19971211122953", "", "text/html", "200", "XAHDNHZ5P3GSSSNJ3DMEOJF7BMCCPZR3", "1405"],
 ["org,archive)/", "19971211122953", "", "text/html", "200", "XAHDNHZ5P3GSSSNJ3DMEOJF7BMCCPZR3", "1405"],
 ["org,archive)/", "19980109140106", "", "text/html", "200", "XAHDNHZ5P3GSSSNJ3DMEOJF7BMCCPZR3", "1402"],


There is some work on custom counters to enchance the aggregation capabilities of CDX server. These features are brand new and should be considered experimental.

Duplicate Counter

While collapsing allows for filtering out adjacent results that are duplicates, it is also possible to track duplicates throughout the cdx using a special new extension. By adding the showDupeCount=true a new dupecount column will be added to the results.

Skip Counter

It is possible to track how many CDX lines were skipped due to Filtering and Collapsing by adding the special skipcount counter with showSkipCount=true. An optional endtimestamp count can also be used to print the timestamp of the last capture by adding lastSkipTimestamp=true

Pagination API

The above resume key allows for sequential querying of CDX data. However, in some cases where very large querying is needed (for example domain query), it may be useful to perform queries in parallel and also estimate the total size of the query.

wayback and cdx-server support a secondary loading from a 'zipnum' CDX index. This index contains CDX lines stored in concatenated GZIP blocks (usually 3,000 lines each) and a secondary index which provides binary search to the 'zipnum' blocks. By using the secondary index, it is possible to estimate the total size of a query and also break up the query in size. Using the zipnum format or other secondary index is needed to support pagination.

However, pagination can only work on a single index at a time; merging input from multiple sources (plain cdx or zipnum) is not possible. As such, the results from a paginated query may be slightly less up-to-date than a default non-paginated query.

  • To use pagination, simply add the page=i param to the query to return the i-th page. If pagination is not supported, the CDX server will return a 400.

  • Pages are numbered from 0 to num pages - 1. If i<0, pages are not used. If i>=num pages, no results are returned.

    Ex: First page:

    Ex: Next Page:

  • To determine the number of pages, add the showNumPages=true param. This is a special query that will return a single number indicating the number of pages


  • Page size is the number of zipnum blocks scanned per page (so a page size of 1 will contain up to 3,000 results per page). This means the number of results on each page will vary, because each block may have a different number of CDX lines matching your query. Page size is configured to an optimal value on the CDX server, and may be similar to max query limit in non-paged mode. The CDX server on currently has a page size of 50.

  • It is possible to adjust the page size to a smaller value than the default by setting the pageSize=P where 1 <= P <= default page size.

    Ex: Get # of pages with smallest page size:

    Ex: Get first page with smallest page size:

  • If there is only one page, adding the page=0 param will return the same results as without setting a page.

  • It is also possible to have the CDX server return the raw secondary index, by specifying showPagedIndex=true. This query returns the secondary index instead of the CDX results and may be subject to access restrictions.

  • All other params, including the resumeKey= should work in conjunction with pagination.

Access Control

The cdx server is designed to improve access to archived data to a broad audience, but it may be necessary to restrict certain parts of the cdx.

The cdx server provides greanting permissions to restricted data via an API key that is passed in as a cookie.

Currently two restrictions/permission types are supported:

  • Access to certain urls which are considered private. When restricted, only public urls are included in query results and access to secondary index is restricted.

  • Access to certain fields, such as filename in the CDX. When restricted, the cdx results contain only public fields.

To allow access, the API key cookie must be explicitly set on the client, eg:

curl -H "Cookie: cdx-auth-token=API-Key-Secret http://mycdxserver/search/cdx?url=..."

The API-Key-Secret can be set in the cdx server configuration.

CDX Server Configuration


Start by editing the wayback-cdx-server-servlet.xml File in the WEB-INF Directory. Just put some valid CDX-Files in the cdxUris-List (Files must end with cdx or cdx.gz!)