Copilot usage metrics dashboard and API in public preview

[ebndev]

The AI era of software development isn’t about are teams using AI, but rather how well are they using it? The GitHub Copilot usage metrics dashboard and API, now in public preview for GitHub Enterprise customers using Copilot, helps you answer that question with data.

You can now see how Copilot adoption and engagement in the IDE varies across both your enterprise and your users to understand who’s using it, where, and how.

What’s included in the public preview

The Copilot usage metrics dashboard gives enterprise administrators and billing managers direct visibility into Copilot usage and adoption under the Insights tab.
The dashboard provides an aggregated enterprise view, while the NDJSON download or API offer user-level granularity for deeper analysis.

In this public preview, you’ll find metrics including:

• Daily and weekly active users across GitHub Copilot IDE modes, including agent mode.
• Agent adoption across your user base.
• Lines of code added and deleted across all IDE modes.
• Language and model usage patterns across IDEs.

These metrics help you understand:

• Overall usage and adoption: Review indicators like weekly usage provide a broad view of Copilot adoption across your enterprise.
• Specific model, feature, and language usage: See which AI models and programming languages are most utilized by your teams, highlighting areas for even greater value.
• Agent adoption percentage: Track how many developers are using Copilot for advanced tasks like refactoring, debugging, and complex problem solving. High agent adoption signals a shift toward truly transformative coding.

Together, these are foundational metrics showing who’s adopting Copilot, how consistently, and what that usage looks like across your enterprise.

How to enable Copilot usage metrics in your enterprise

By default, Copilot usage metrics are disabled unless your enterprise participated in the private preview, in which case they remain enabled.

Enterprise admins can enable them as follows:

1. Navigate to your AI Controls tab on github.com.
2. Select Copilot in the left sidebar.
3. Scroll down to the "Metrics" section where you’ll see Copilot usage metrics.
4. Change the Copilot usage metrics policy setting to Enabled. This will enable the Copilot usage metrics dashboard and API for your enterprise, allowing enterprise administrators and billing managers to view usage and adoption data.

Once enabled, you can access the Copilot usage metrics dashboard directly from the Insights tab at the enterprise level or through the Enterprise and user-level API.

Learn more

• View the Copilot usage metrics dashboard documentation.
• Explore the Copilot usage metrics API.
• Join us in the comments below to share your feedback and ask questions.

Disclaimer: The UI for features in public preview is subject to change.

[yu-liang-dev]

Hi team, is there any documentation available yet for the user-level API discussed in the article? Thanks in advance 🙏

I am still only seeing the organization/team-level APIs in the REST API documentation, and a search yielding 20 "Copilot" results did not contain any information about the user-level API.

[alexlato22]

Hi @yu-liang-dev -- yes, you can find it here: https://docs.github.com/en/enterprise-cloud@latest/rest/copilot/copilot-metrics?apiVersion=2022-11-28#get-copilot-metrics-for-an-enterprise

under these endpoints:

get/enterprises/{enterprise}/copilot/metrics/reports/users-1-day
get/enterprises/{enterprise}/copilot/metrics/reports/enterprise-28-day/latest
get/enterprises/{enterprise}/copilot/metrics/reports/enterprise-1-day
get/enterprises/{enterprise}/copilot/metrics/reports/users-28-day/latest

[yu-liang-dev]

@alexlato22 , thanks for the quick turnaround!

[warrenandre]

Hi, Could you please update documentation to include enterprise usage metrics as I am not finding it on the documentation. there are only links to the metrics that are available but no API examples for Enterprise usage Metrics. the page has a link to Copilot usage metrics API. but no endpoint information is available.

[alexlato22]

Hi @warrenandre. We are working on improving the docs as we speak. Sorry about the confusion. Here are the 4 endpoints to look at for Usage metrics specifically on that docs page:

get/enterprises/{enterprise}/copilot/metrics/reports/users-1-day
get/enterprises/{enterprise}/copilot/metrics/reports/enterprise-28-day/latest
get/enterprises/{enterprise}/copilot/metrics/reports/enterprise-1-day
get/enterprises/{enterprise}/copilot/metrics/reports/users-28-day/latest

[svc-git-integration]

Hey, could you also provide some more info about the authentication token required to hit the endpoint
And a few examples on how to handle the response will be really helpful.

[warrenandre]

Thank you, @alexlato22 I appreciate it.

[alexlato22]

November 14th Update (Data refinement):

Update: Improving Accuracy of Copilot Chat Metrics

As Copilot’s chat, agent, and editing experiences expand, we’re refining how activity is identified so your dashboards reflect real user behavior with greater accuracy.

Some IDE telemetry represent general communication between the editor and the Copilot server. These happen automatically and aren’t always tied to a specific feature or user action. Since they aren’t meaningful indicators of chat engagement, we’re updating our classification so only true chat interactions appear in chat specific metrics.

What’s changing

• Generic client to server requests will no longer be interpreted as chat activity
• These signals will still enrich model and mode information when they occur alongside real interactions
• Activity counts will now reflect only clear, user initiated behavior

What you may notice

• More accurate “used chat” counts
• A major reduction in “unknown” chat modes
• A small decrease in Daily Active Users in cases where only these automatic IDE signals were present
• Removal of a few internal to GitHub model identifiers that previously appeared

This refinement makes the metrics easier to interpret and keeps them aligned with how your users actually engage with Copilot as the product surface grows.

If you have questions or want help interpreting the updated numbers, feel free to reply here. Thank you!

[warrenandre]

Yes, I am willing to assist with interpreting the new updated numbers.

[alexlato22]

Updated Model Visibility in Your Copilot Metrics

You may have noticed a short-term increase in “Unknown” model entries in your Copilot metrics last week or in the recent weeks. This happened because several newly released models became available before our reporting system had their names mapped. Usage and billing were never affected; this was purely a labeling delay.

We’ve now updated the pipeline to fully recognise all the new models:

Newly recognized models:

• GPT 5.1
• GPT 5.1 Codex
• GPT 5.1 Codex Mini
• GPT 5 Codex
• Claude Haiku 4.5
• Claude Sonnet 4.5
• Claude Opus 4.1 (all variations)
• Gemini 3.0 Pro
• Gemini 2.5 Flash (improved label)

Your dashboards and API responses will now show these correctly going forward.

We’re also strengthening our process so we get earlier notice of upcoming model releases which will allow us to prepare mappings ahead of time and keep your data accurate from day one.

If anything looks off or you have questions about a specific model, we’re here to help. ❤️

[dy-resource]

Will the metric loc_added_sum be made available for the organization plan API in the future? Currently, the organization API only allows access to the LOC metric for code completion, excluding that for the agent/edit mode.

[alexlato22]

Yes! We are working on adding an org endpoint now and hope to have it ready in the next few weeks.

[alexlato22]

Copilot Metrics 'Code generation' Dashboard – Public Preview Rollout 🚀

We've released a new Code generation Metrics Dashboard into public preview. This dashboard is the interactive companion to the API lines of code metrics shipped in September and brings LoC data directly into a visual experience for you.

If you are in public preview you will already see Copilot usage and will now automatically get Code generation.

Enabling the dashboard

The dashboard is opt in at the enterprise level.

Customers can enable it from:

AI Controls → Copilot → Metrics

Once enabled, the dashboard appears immediately and API access is unlocked. If you are already opted in, the new dashboard will automatically appear.

Refresher on Lines of Code Metric

A quick reminder on how LoC metrics are calculated and feel free to read the updated docs here.

Core fields

• loc_suggested_to_add_sum – lines Copilot suggested to add
• loc_suggested_to_delete_sum – lines Copilot suggested to delete
• loc_added_sum – lines actually added to the editor
• loc_deleted_sum – lines actually deleted from the editor
• agent_edit – bucket capturing agent and edit mode changes (added + deleted)

User initiated vs agent initiated

User initiated

Completions, inline chat, chat panel ask, edit, custom, and unknown modes.

Agent initiated

Only agent_edit.

Agent mode behaves differently. There’s no suggestion → acceptance event. The agent plans a task, edits multiple files, retries, and self corrects as it works. Edits land straight in the editor, so there’s no “suggested line count.” Counting added + deleted lines gives a clearer picture of the actual impact, especially on larger or multi file changes.

Key Dashboard Charts

1. Total Lines Changed with AI

Shows Copilot’s total code generation volume by adding up all added and deleted lines across all features.

2. Agent Contribution

Highlights how much of the total code change came from agent mode.
Formula: agent_edit (added + deleted) / all added + deleted lines

3. Average Lines Deleted by Agent

Shows how much cleanup agent mode performs per user.
Formula: agent deleted lines / number of users who used agent mode

4. User vs Agent Code Changes

• User initiated: suggested vs added
• Agent: added vs deleted (no suggestions for agent edits)

5. Breakdowns by model and language

Pattern holds:

• User initiated: suggested + added
• Agent: added + deleted
• Unknown is excluded at the model level to match the usage dashboard.

Please let us know here if you have any questions or feedback, thank you!

[lvthillo]

Any plans to implement this at the organizational level as well?