How it works

From a GitHub organization to a team dashboard, in four steps and under ten minutes.

Here is exactly what Squadlog connects to, what it reads, what it keeps, what it writes, and what it never does. No marketing, the actual mechanics.

1Connect

Install a read-only GitHub App.

Squadlog is a GitHub App, not a personal token. You install it on one or several GitHub.com organizations, or on a GitHub Enterprise Server 3.12+ instance, and choose which repositories it may see. GitHub then pushes events to Squadlog as they happen: commits, pull requests, reviews, merges.

  • Read access to repository metadata, commits and pull requests
  • Read access to the members of the organization
  • No write access, anywhere. No access to issues, discussions or secrets.
  • Webhooks are verified by signature; a connection that stops sending events is flagged within the hour.
GitHub connections2 organizations · 1 GHES
Aacmegithub.com · GitHub App · 14 repos accessibleConnected
Gghes.acme.internalGitHub Enterprise Server 3.13 · 22 reposConnected
Lacme-labsgithub.com · no event for 3hChecking
2Create a team

A name, a color, its repos, its managers.

A Squadlog team is yours, not GitHub’s. Pick the repositories it works on across any connected organization, name its managers, and choose its primary measure: merged pull requests for teams that review everything, commits for teams that push to main. Both are always tracked.

  • One repository can belong to several teams; one person can be in several teams
  • History loads in the background: 12 months
  • Thresholds, working days and time zone are set per team and can change any time
Team · Platformprimary measure: commits
onacme/api-gatewayalso tracked by Payments31 commits / 30d
onacme/webalso tracked by Web12 commits / 30d
onacme/infraonly Platform5 commits / 30d
offacme/design-systemnot tracked—
3Members

The team fills itself. Nobody signs up.

Everyone who contributes to a tracked repository appears as a member. Several GitHub logins can be merged into one person, across organizations. Each linked member receives an information email describing what is tracked, what is not, and their rights. Developers never need a Squadlog account.

  • Logins detected from commits and PRs, matched to people by the manager
  • Declared absences pause the inactivity alert and are excluded from working-day counts
  • Effort scores and rankings seen by managers only, no automated decision
Members · 8linked from 2 connections
KOKofi Oseik-osei · github.com + kofi.osei · ghesActive
MLMarie LefèvremlefevreActive
TRTomás Ruiztruiz · 2 PRs pendingWatch
SVSanna Virtanensvirtanen · parental leave Sep 1 → Nov 30On leave
4Read and summarize

Each commit is read once, then forgotten.

When a commit lands or a PR merges, Squadlog loads the diff in memory, has it read by a model running on its own servers, writes one line a non-technical reader understands and scores the effort from 1 to 10. The summary and the score are kept. The diff is discarded. That is the only moment Squadlog touches code.

  • Own models, own servers: nothing goes to a third-party AI provider, nothing is used for training
  • Excluded repositories are never read, even at merge; members can opt out
  • Everything else (dashboards, alerts, reports) is computed from metadata only
What happens to a commita few seconds after it lands
1Commit landsacme/api-gateway · k-osei · +612 −248webhook
2Diff read once, in memorynever written to diskdiscarded
3Our model reads the intentruns on Squadlog serversno training
4Only this is keptQueue incoming GitHub events with retries.9/10
5 · What is kept, what is not

The whole data story on one card.

read-only GitHub App · own models · no training
Kept✓Author, date, repository, branch name of each commit and PR✓Lines added and removed✓PR and review events: opened, reviewed, approved, merged✓One-line summary per commit and per PR✓Effort score 1 to 10 per commit✓Team aggregates for 12 months
Never kept×File contents×Diffs, once read×Issues, discussions, comments×Secrets or environment variables×Anything for training a model×Any individual rating of a person
Read once, in memoryWhen a commit lands or a PR merges, its diff is loaded, read by a model running on Squadlog's own servers, summarized in one line and scored. The diff is then discarded. Nothing is sent to a third-party AI provider and nothing is used for training.Hosted in Paris · DPA available · GDPR art. 22 compatible
6 · A week with Squadlog

What a manager actually sees, day by day.

Monday 08:00Digest and weekly reportThe week in plain words, three figures, the highlights, the PDF attached. Read it before stand-up.
TuesdayAn alert opensA PR on the gateway has waited two working days without review. One email, a push on the phone. You nudge the reviewer.
WednesdayA look at the timelineFilter by Tomás: nine commits this week, the rate limiter scored 8/10 yesterday. Nothing to do, good to know.
ThursdayThe alert closes itselfThe PR was reviewed and merged. The alert resolves on its own and says so in Notifications.
FridayPrepare the 1:1Open Marie’s profile: her commits in plain words, the themes she carried, her reviews given. Facts and effort scores to talk about.

Ten minutes from now, your first team is on the dashboard.

7 days free, every feature. No credit card. Developers never need an account.