OmniscioGuidesClaude Code analytics
Claude Code analytics: team usage in one dashboard
Every team that buys Claude Code eventually has to answer the same question — is it being used, and is the work better? The analytics dashboard is where that answer is supposed to live, and it is easy to read the wrong thing into it.
This page covers which dashboard your plan gets, how to turn contribution metrics on, exactly what each number counts, and the files and rewrites the attribution deliberately ignores.
Last checked against Anthropic's Claude Code documentation on November 18, 2026. By the Omniscio team. Updated .
Which dashboard your plan gets
- Claude for Teams or Enterprise:
claude.ai/analytics/claude-code. Usage metrics, contribution metrics with GitHub integration, a leaderboard, and CSV export. - API (Claude Console):
platform.claude.com/claude-code. Usage metrics, spend tracking and team insights.
They are not the same dashboard with a different logo. Contribution metrics — the PR and line-of-code numbers — exist only on the Team and Enterprise side, and only when GitHub is wired in.
Source: the plan table at the top of Anthropic's analytics page, read November 18, 2026.
The Team and Enterprise dashboard
Admins and Owners see four sections:
- Usage metrics: lines of code accepted, suggestion accept rate, daily active users and sessions.
- Contribution metrics: PRs and lines of code shipped with Claude Code assistance, with GitHub integration.
- Leaderboard: top contributors ranked by Claude Code usage.
- Data export: contribution data as a CSV for your own reporting.
There is a gap worth knowing before someone files a ticket about it: the dashboard does not show per-user token counts or cost estimates. For those, configure OpenTelemetry export, or export the spend report from your organization's analytics settings, which lists token usage and estimated usage-credit spend per user and per model.
Source: "Access analytics for Team and Enterprise" in Anthropic's analytics page, read November 18, 2026.
Turning on contribution metrics
Usage and adoption data is there for every Team and Enterprise account. Contribution metrics are a public beta that needs a GitHub organization connected, and the setup is four steps split between two people:
- A GitHub admin installs the Claude GitHub app on the organization's account.
- A Claude Owner enables the Claude Code analytics feature at
claude.ai/admin-settings/claude-code. - The same Owner enables the GitHub analytics toggle on that page.
- Someone completes the GitHub authentication flow and selects which organizations to include.
Then wait: data typically appears within 24 hours and updates daily. Two messages explain a blank panel — "GitHub app required" means the app is not installed, and "Data processing in progress" means this particular step can take a few days, so check back and confirm the app is installed if it stays blank.
Two limits decide whether it can work at all. Organizations with Zero Data Retention enabled never see contribution metrics; the dashboard shows usage only. And the metrics cover repositories on github.com: for GitHub Enterprise Server, again usage only.
Source: "Enable contribution metrics" in Anthropic's analytics page, read November 18, 2026.
What the summary numbers count
Anthropic states the caveat before the definitions, and it is the important part: these metrics are deliberately conservative and represent an underestimate of Claude Code's actual impact, because only lines and PRs with high confidence in Claude Code's involvement are counted.
- PRs with Claude Code: merged pull requests containing at least one line written with Claude Code. One card shows the count, another the share of all merged PRs.
- LOC with Claude Code: total lines across all merged PRs that were written with Claude Code assistance. Only "effective lines" count — lines with more than 3 characters after normalization.
- Suggestion accept rate: the percentage of times users accept Claude Code's editing suggestions, including Edit, Write and NotebookEdit tool use.
- Lines of code accepted: total lines the Edit and Write tools added or removed in accepted edits. Rejected suggestions do not count, and lines deleted later do not reduce the total.
Read together, those four answer different questions: adoption (accept rate, active users), output (lines accepted), and shipped work (PRs and LOC with Claude Code). A team with a high accept rate and flat PRs is using it and not shipping more, which is a different problem from a team that is not using it.
One Claude line is enough to count a whole PR, which sounds generous against the "underestimate" framing. Both are true because the matching is strict at the line level: only lines it can tie to a session with high confidence are credited, and generated files, lock files, build output and code rewritten by more than 20% are not credited at all. A PR with a single surviving Claude line counts; a PR the algorithm cannot prove was not counted.
Source: "Review summary metrics" in Anthropic's analytics page, read November 18, 2026.
How a PR gets attributed, and what it skips
Attribution happens when a pull request is merged. Claude Code extracts the added lines, finds sessions that edited matching files inside a time window, matches the two with several strategies, and labels the PR claude-code-assisted in GitHub.
The window is 21 days before to 2 days after the merge date. Before comparing, lines are normalized: whitespace trimmed, repeated spaces collapsed, quotes standardized, text lowercased.
Some things are excluded by design, which explains a low number in a repository that looks heavily AI-assisted:
- Lock files such as
package-lock.json,yarn.lockandCargo.lock. - Generated code — Protobuf output, build artifacts, minified files.
- Build directories such as
dist/,build/,node_modules/andtarget/. - Test fixtures — snapshots, cassettes, mock data.
- Lines over 1,000 characters, likely minified or generated.
And code that developers substantially rewrote — more than 20% different — is not attributed at all. Nor is the algorithm influenced by which branch a PR came from or went to.
Source: "PR attribution" in Anthropic's analytics page, read November 18, 2026.
For API customers
The Console dashboard at platform.claude.com/claude-code needs the UsageView permission. Developer, Billing, Admin, Owner and Primary Owner roles have it.
It shows lines of code accepted, suggestion accept rate, daily active users and sessions, and daily API spend in dollars. The team insights table adds per-user spend and per-user lines accepted for the date range at the top of the page, the last 30 days by default; API-key users appear by key name and OAuth users by email address.
Two things it does not do. Contribution metrics with GitHub integration are not available to API customers — usage and spend only. And the spend figures are estimates for analytics purposes; the Cost page is where actual costs live.
Sources: "Access analytics for API customers" and "View team insights" in Anthropic's analytics page, read November 18, 2026.
Getting the data out
Three roads, depending on what you have:
- Enterprise Analytics API. On the Enterprise plan, it returns per-user engagement, usage and cost reports across Claude surfaces including Claude Code. A Primary Owner creates a key with the
read:analyticsscope. It is not available on the Teams plan. - Claude Code Analytics API. API customers pull daily per-user metrics with an Admin API key.
- The CSV export and the GitHub label. Export all users downloads full contribution data, not just the top ten on the leaderboard, and searching for PRs labeled
claude-code-assistedgives you the raw list from GitHub instead.
Source: "Access data programmatically" in Anthropic's analytics page, read November 18, 2026.
What turns the data off
Two switches, and they are easy to trip by accident. Zero Data Retention removes contribution metrics entirely, as above. And delivering DISABLE_TELEMETRY through managed settings stops the usage data that feeds this dashboard for every developer the policy reaches — which is the right call for some organizations and a silent reason for an empty dashboard in others.
Source: "Turn telemetry off for your organization" in Anthropic's managed settings page, read November 18, 2026.
Where these answers come from
- Everything on this page: Track team usage with analytics, read November 18, 2026.
- Real-time metrics for your own stack: Monitoring with OpenTelemetry.
- What one person's own
/usageshows: token usage. - The policy switch that stops the feed: Deploy managed settings.
Common questions
Where is the Claude Code analytics dashboard?
Teams and Enterprise: claude.ai/analytics/claude-code, for admins and Owners. API customers: platform.claude.com/claude-code.
Why does my dashboard show usage but no contribution metrics?
The beta needs the Claude GitHub app installed and both analytics toggles on. Zero Data Retention removes it entirely, and GitHub Enterprise Server repositories show usage only.
How long after a pull request merges does it count?
Attribution looks at sessions from 21 days before to 2 days after the merge date, then labels the PR claude-code-assisted.
Are the numbers an overestimate?
No — Anthropic calls them deliberately conservative and an underestimate, since only high-confidence lines and PRs count.