Grep Origin Contents via the API

Search the text of a repository's files at a ref with POST https://api.cursor.com/v1/origin/repos/{ownerSlug}/{repoName}:grep. Official reference: Origin API and changelog (OpenAPI OriginService_GrepContents).

Origin is in Early Beta and subject to change. Authenticate with an installation access token (oit_…) or a user access token that carries repository:contents:read. Availability: Early Beta per the API overview.

The body requires query. By default query is a regular expression; set literal to true to match exact text. The search is line-oriented: a pattern never matches across a line break, and each returned entry is one line. There is no pagination; the response is complete only when limitHit is false.

Cost: 5 points against the principal budget (heavier read, same tier as Get Commit and Get Repo Tarball).

Path parameters

Param	Required	Meaning
ownerSlug	yes	Owning entity's unique slug
repoName	yes	Repo name, unique to the owner

Request body

Field	Required	Meaning
query	yes	Pattern to search. Empty is rejected (400). Max UTF-8 size: 4096 bytes
ref	no	Commit, branch, tag, or symbolic ref such as HEAD. Empty uses the default branch
literal	no	When true, treat query as exact text instead of a regex
caseInsensitive	no	Applied only when literal is true; ignored for regex (use a leading (?i) in query)
wholeWord	no	Applied only when literal is true; ignored for regex (use \b around the pattern)
contextBefore	no	Lines of context before each match; values above 10 reduce to 10
contextAfter	no	Lines of context after each match; values above 10 reduce to 10
filterPath	no	Restrict to this file or directory relative to the repo root
includes	no	Glob allowlist (at most 20). Case-insensitive; * is one segment; ** crosses segments
excludes	no	Glob denylist (at most 20). An exclude beats an include
maxResults	no	Cap on matching occurrences. 0 or omit defaults to 1000; values above 1000 reduce to 1000. Context lines do not count

curl --request POST \
  --url 'https://api.cursor.com/v1/origin/repos/OWNER_SLUG/REPO_NAME:grep' \
  --header 'Authorization: Bearer YOUR_ORIGIN_TOKEN' \
  --header 'Content-Type: application/json' \
  --data '{
  "query": "(?i)launch",
  "ref": "HEAD",
  "includes": ["**/*.ts"],
  "excludes": ["**/node_modules/**"],
  "contextBefore": 1,
  "contextAfter": 1,
  "maxResults": 100
}'

Literal whole-word example:

curl --request POST \
  --url 'https://api.cursor.com/v1/origin/repos/OWNER_SLUG/REPO_NAME:grep' \
  --header 'Authorization: Bearer YOUR_ORIGIN_TOKEN' \
  --header 'Content-Type: application/json' \
  --data '{
  "query": "launch",
  "literal": true,
  "caseInsensitive": true,
  "wholeWord": true
}'

Response

200 returns matches (order of files and lines is unspecified) and limitHit.

Each entry is a GrepContentsMatch:

Field	Meaning
path	Path relative to the repository root
lineNumber	One-based line number
line	Line text without its trailing terminator
kind	match or context
submatches	Byte ranges inside line (start inclusive, end exclusive). Always empty on context lines

{
  "matches": [
    {
      "path": "src/cli.ts",
      "lineNumber": 42,
      "line": "  await launchAgent(opts);",
      "kind": "match",
      "submatches": [
        {
          "start": 8,
          "end": 14
        }
      ]
    }
  ],
  "limitHit": false
}

An empty repository (no refs) returns no matches and limitHit false. When limitHit is true, narrow query, filterPath, or the glob lists. Base URL: https://api.cursor.com/v1/origin.