Create Commit from Files

Create a commit by streaming file operations directly to the service. This is designed for automation, AI agents, and other workflows that need atomic Git writes without running a local Git process.

JWT claims

Required scopes: git:write
Requires per repo scope: Yes

POST /api/repos/{repo_name}/commit-pack

Create Commit from Files

cURL

curl --request POST \
  --url https://api.{cluster}.code.storage/api/repos/{repo_name}/commit-pack \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/x-ndjson' \
  --data '
"{\"metadata\":{\"target_branch\":\"main\",\"expected_target_sha\":\"d34db33fd34db33fd34db33fd34db33fd34db33f\",\"commit_message\":\"Update docs\",\"author\":{\"name\":\"Docs Bot\",\"email\":\"[e-mail]\"},\"files\":[{\"path\":\"docs/changelog.md\",\"operation\":\"upsert\",\"content_id\":\"blob-1\",\"mode\":\"100644\"},{\"path\":\"docs/legacy.txt\",\"operation\":\"delete\",\"content_id\":\"blob-2\"}]}}\n{\"blob_chunk\":{\"content_id\":\"blob-1\",\"data\":\"IyBWZXJzaW9uIDIuMS4wCi0gcmVmcmVzaCBkb2NzCg==\",\"eof\":true}}\n{\"blob_chunk\":{\"content_id\":\"blob-2\",\"data\":\"\",\"eof\":true}}\n"
'

201

{
  "commit": {
    "blob_count": 1,
    "commit_sha": "<string>",
    "pack_bytes": 1,
    "target_branch": "<string>",
    "tree_sha": "<string>"
  },
  "result": {
    "branch": "<string>",
    "new_sha": "<string>",
    "old_sha": "<string>",
    "status": "<string>",
    "success": true,
    "target_branch": "<string>",
    "message": "<string>"
  }
}

Authorizations

Authorization · string · header · required

JWT bearer token signed with your organization's registered signing key. There is no OAuth authorization server; you mint tokens yourself and the token's scopes claim carries the granted permission scopes. Public scopes: git:read (read repository contents), git:write (write branches, commits, tags, and notes), repo:write (create and manage repositories), org:read (list an organization's repositories). Each operation's security requirement names the scopes it requires. See https://code.storage/docs/getting-started/authentication for how to mint tokens.

Path Parameters

repo_name · string · required

Repository name. Names that contain / or any other character that is not safe in a URL path segment must be URL encoded so the value occupies a single path segment. For example pierre/example is sent as pierre%2Fexample. Plain names such as example can be sent as-is. The server URL-decodes the value before resolving the repository.

Body

application/x-ndjson

Send newline-delimited JSON where the first line is a metadata object and each following line is a blob_chunk. This lets you stream file contents without a local Git client.

NDJSON stream. The first line must be metadata; subsequent lines must be blob_chunk payloads. Decoded blob chunks are limited to 4 MiB each.

Response

201

application/json

Commit created and written to the target branch.

Result of applying an NDJSON commit pack, including the resulting commit and branch update.

commit · object · required

Commit object information generated by the storage service.

Show child attributes

result · object · required

Final branch update outcome.

Show child attributes