Docs
Connections

Git overview

Chart repositories, commits, pull requests, issues, releases, tags, branches, contributors and pipeline runs from GitHub, GitLab, Gitea, Bitbucket and Azure DevOps, read-only, in one set of columns.

A Git connection reads analytics data from a hosted code forge with a read-only token. Every forge reads into the same tables and columns, so a chart of commits per week or pull request lead time works the same on GitHub, GitLab, Gitea, Bitbucket or Azure DevOps.

Preview

Git connections are in Preview, and the Git tile in New connection carries a Preview badge. They are not enabled by default on xentropy.ai: contact support@xentropy.ai to request access. Until a forge is enabled for your account it is not offered in the forge selector, and its reads fail with PROVIDER_DISABLED.

What you can chart

  • Delivery: commits per week, by repository or author; pull requests opened, merged and closed; time from open to merge.
  • Work tracking: issues (Azure DevOps work items) opened and closed, by type, label or assignee.
  • Releases: release cadence, tags and pre-releases over time.
  • CI health: pipeline runs by status and conclusion, run duration, failure rate per branch.
  • Inventory: repositories with their stars, forks, open issues, default branch and last push; branches and protected branches.

Forges and hosts

Git connections reach only the hosted platforms below. Self-hosted servers (GitHub Enterprise Server, self-managed GitLab, your own Gitea or Forgejo instance, Azure DevOps Server) are not supported yet, and the form has no server address to enter.

ForgeHost Xentropy callsCredential
GitHubapi.github.comFine-grained personal access token
GitLabgitlab.comPersonal, group or project access token with read_api
Giteagitea.comAccess token with read permissions
Bitbucketapi.bitbucket.orgAtlassian API token with scopes, or an access token
Azure DevOpsdev.azure.comOrganization-scoped personal access token

Each forge's connection can call only the read endpoints Xentropy needs on that forge's own host, so no IP allowlist is needed.

Read-only by design

  • Every guide asks for read-only permissions only. Xentropy never asks for write or admin scopes.
  • Xentropy sends only read requests. The one exception is Azure DevOps work items, which are found with a WIQL query (a read sent as POST).
  • Tokens are encrypted at rest, sent only in request headers to the forge's API, and never shown again after you save them.
  • Redirects are never followed. A repository that moved or was renamed fails with NOT_FOUND until you select its new name.
  • Xentropy reads on a schedule or on demand. It does not install webhooks.

Connect a forge

Create a read-only token on the forge. Each forge's guide lists the exact permissions.
Open Connections, choose New connection, then Git, and choose the forge.
Name the connection and paste the token. Azure DevOps also needs the organization; Bitbucket takes an optional workspace and, for an Atlassian API token, the account e-mail.
Choose Test connection. The test checks Authentication (the token) and Repository access, and lists the repositories the token can read.
Choose the repositories to allow, or none to allow every repository the token can read (up to 100 repositories per connection). Then choose Save connection.
Open the connection, choose a resource, the repositories and, for history, the date range. Pick fields, preview 100 rows and save a dataset.

For GitHub, GitLab and Bitbucket, a Connect with … sign-in button may appear above the token fields. It appears only when sign-in with that forge is available for your account; otherwise use a token.

Dataset options

OptionApplies toDefault
ResourceAll—
RepositoriesEverything except repositoriesRequired: 1 to 20 repositories per dataset, from the connection's selection
BranchcommitsThe default branch
Date range (since, until)commits, pull_requests, issues, pipeline_runsThe last 90 days
Statepull_requests (open, closed, merged), issues (open, closed)All

Commits are filtered on committed_at; pull requests, issues and pipeline runs on created_at. The range includes since and excludes until. closed never includes merged pull requests, so open, closed and merged split the set on every forge.

Tables and columns

Every table except repositories starts with repository, the repository's full name (owner/name; for Azure DevOps project/repository). Types are string, number, boolean and datetime (UTC). A column the forge cannot fill is empty (null); each forge's guide lists them under Columns left empty. extra holds the forge's own fields as a JSON object. It is off by default.

repositories

One row per repository the connection can read.

ColumnTypeNotes
idstringThe forge's id
full_namestringRow key
namestring
ownerstringOwner, group, workspace or project
descriptionstring
visibilitystringpublic, private or internal
default_branchstring
forkboolean
archivedboolean
languagestring
starsnumber
forksnumber
open_issuesnumber
size_kbnumber
created_atdatetime
updated_atdatetime
pushed_atdatetimeLast push
web_urlstring
extrastringJSON object, off by default

commits

One row per commit on the branch, in the date range.

ColumnTypeNotes
repositorystring
shastringRow key, with repository
messagestringFirst line
author_loginstringThe author's forge handle
author_namestringPersonal data
author_emailstringPersonal data
authored_atdatetime
committer_namestringPersonal data
committer_emailstringPersonal data
committed_atdatetimeDate range column
parentsnumber2 or more for merge commits
additionsnumberLines added
deletionsnumberLines deleted
web_urlstring
extrastringJSON object, off by default

pull_requests

Pull requests (GitLab merge requests), created in the date range.

ColumnTypeNotes
repositorystring
numbernumberRow key, with repository
idstring
titlestring
statestringopen, closed or merged
draftboolean
author_loginstring
source_branchstring
target_branchstring
created_atdatetimeDate range column
updated_atdatetime
closed_atdatetime
merged_atdatetime
additionsnumberLines added
deletionsnumberLines deleted
changed_filesnumber
commentsnumber
web_urlstring
extrastringJSON object, off by default

issues

Issues (Azure DevOps work items), created in the date range. Pull requests are never counted as issues.

ColumnTypeNotes
repositorystring
numbernumberRow key, with repository
idstring
titlestring
statestringopen or closed
typestringissue, or the Azure DevOps work item type
author_loginstring
assignee_loginstring
labelsstringJSON list
created_atdatetimeDate range column
updated_atdatetime
closed_atdatetime
commentsnumber
web_urlstring
extrastringJSON object, off by default

releases

ColumnTypeNotes
repositorystring
idstringRow key, with repository
tagstring
namestring
draftboolean
prereleaseboolean
author_loginstring
created_atdatetime
published_atdatetime
web_urlstring
extrastringJSON object, off by default

tags

ColumnTypeNotes
repositorystring
namestringRow key, with repository
commit_shastringThe tagged commit
tagged_atdatetime
extrastringJSON object, off by default

branches

ColumnTypeNotes
repositorystring
namestringRow key, with repository
commit_shastringHead commit
protectedboolean
defaultbooleanThe repository's default branch
extrastringJSON object, off by default

contributors

ColumnTypeNotes
repositorystring
loginstringRow key, with repository
namestringPersonal data
emailstringPersonal data
contributionsnumberCommits: all time where the forge counts them, or in the date range when counted from commits
extrastringJSON object, off by default

pipeline_runs

CI runs (GitHub and Gitea Actions, GitLab pipelines, Bitbucket Pipelines, Azure Pipelines builds), created in the date range.

ColumnTypeNotes
repositorystring
idstringRow key, with repository
namestringPipeline or workflow
run_numbernumber
statusstringqueued, in_progress or completed
conclusionstringsuccess, failure, cancelled, skipped, or empty while running
branchstring
commit_shastring
eventstringWhat triggered the run
actor_loginstringWho triggered it
created_atdatetimeDate range column
started_atdatetime
finished_atdatetime
duration_snumberSeconds
web_urlstring
extrastringJSON object, off by default

Support by forge

Yes means the forge's API has the resource. From commits means Xentropy counts contributors from the commits in the date range (by login, else e-mail, else name), because the forge has no contributors list. No means the forge has no such data: choosing it fails with VALIDATION_FAILED before any request is sent.

ResourceGitHubGitLabGiteaBitbucketAzure DevOps
repositoriesYesYesYesYesYes
commitsYesYesYesYesYes
pull_requestsYesYesYesYesYes
issuesYesYesYesNo (issue tracker API removed)Yes (project work items)
releasesYesYesYesNoNo
tagsYesYesYesYesYes
branchesYesYesYesYesYes
contributorsYesYesFrom commitsFrom commitsFrom commits
pipeline_runsYesYesYesYesYes

Where a forge has no releases, read tags instead.

Personal data

Commit author and committer names and e-mails, and contributor names and e-mails, are marked Personal data in the field picker and are off by default. Logins are public handles and are not marked. If you select a personal data field, the dataset is marked as containing personal data (containsPii). The API and agents must confirm this with containsPii: true, otherwise the selection is refused.

Limits and refresh

Every read is bounded, and a read either returns complete data or fails.

LimitValue
Pages per readFree 100, Pro 600 (shared by all repositories in the dataset)
Rows per page100 on GitHub, GitLab and Azure DevOps; 50 on Gitea and Bitbucket
Rows per read50,000
Data per read5 MiB
Read deadline30 s interactive, 300 s scheduled
Repositories per dataset20 repositories
Repositories per connection100 repositories, or every repository the token can read
History windowThe last 90 days unless you set the range

A history resource on a busy repository can need many pages: on the Free plan, 100 pages of 100 commits is about 10,000 commits per read (5,000 on Gitea and Bitbucket). A read over the limits fails with LIMIT_EXCEEDED or BUDGET_EXCEEDED; narrow the date range or the repositories, or upgrade to Pro.

Rate limits

Xentropy follows each forge's rate limit headers and pauses before the limit is reached. Limits are counted per token, so other tools that share the token share its budget. If the forge asks Xentropy to wait longer than the read's deadline, the read fails with RATE_LIMITED and details.retryAt says when the forge will accept requests again. Scheduled refreshes of that connection wait until then, then run again by themselves. A failed read never replaces earlier data: charts keep their last capture.

Errors

CodeCauseWhat to do
PROVIDER_ACCESS_DENIED, stage authThe forge rejected the token (HTTP 401): expired, revoked or mistypedCreate a new token and choose Replace configuration
PROVIDER_ACCESS_DENIED, stage permissionsA permission is missing (HTTP 403). details.missingScopes names itAdd the permission to the token, or create a new one, then test again
NOT_FOUNDThe repository does not exist, moved or was renamed, or the token cannot see it (HTTP 404 or a redirect)Check the name on the forge, grant the token access, or select the new name
RATE_LIMITEDThe forge's rate limit is used up (HTTP 429, or 403 with rate limit headers)Retry after details.retryAt; scheduled refreshes wait by themselves
VALIDATION_FAILEDThe forge has no such resource, or the options are invalidChoose a supported resource; see the support table
PERMISSION_DENIEDThe repository is not in the connection's selectionAdd it to the connection, or choose a selected repository
INCOMPLETE_DATAThe forge's paging could not be proven complete, or Azure DevOps has 20,000 or more work itemsRetry, or narrow the date range
LIMIT_EXCEEDED, BUDGET_EXCEEDEDMore pages, rows or time than the plan allowsNarrow the date range or repositories, or upgrade to Pro
PROVIDER_UNAVAILABLEThe forge failed (HTTP 5xx) or could not be reachedRetry later
PROVIDER_DISABLEDThe forge is not enabled for your accountContact support

Agents

An agent reads a Git connection with an API key that has connections:read or git:read. Datasets use the git resource: resource, repositories, and optionally branch, since, until, state, fields and containsPii. Creating a connection needs connections:manage, as for every kind; see Connections.

GitHub data and the OpenAPI preset

For GitHub data, use a Git connection: it gives typed tables for repositories, commits, pull requests, issues and workflow runs, with repository selection, personal data off by default and a read-only token. The OpenAPI connection's GitHub preset stays for other GitHub API endpoints and returns their raw JSON. Existing OpenAPI GitHub connections keep working; nothing is migrated.

On this page