In order to enable optimal AI ROI evaluation, GitClear integrates with all the major code AI providers. This page describes how - and what - data can be retrieved to measure the exact impact of the team's AI usage. You can jump straight to your AI provider if you want to get started connecting to GitClear:

Each AI Provider has somewhat different data that is provided via their API. And, like all LLM-adjacent code, these evolve fast. If a provider doesn't return the data you wish today, there's a fair chance they will next quarter. That said, as of Q2 2026, here is the data available from each AI Provider:

Platforms that provide less data to GitClear force a greater magnitude of extrapolation when reconciling the AI Provider-supplied data with the observed commits made by the team. For telemetry-based setups, GitClear can map from exact changed lines to the LLM that changed them, with confidence above 95%. For less robust AI APIs, like Gemini Code Assist and Augment Code, we can still confidently evaluate the total AI adoption rate for the team, and how AI is correlating with various outcomes, like "more tickets closed." However, these low-information APIs reduce the ability to derive a longer-term derivation like "how much more often did the AI write code that went on to create a critical defect?"

linkAI Provider Setup form

All AI Usage APIs, besides the "Claude & Codex telemetry"-based ones, are set up by visiting the rightmost "Settings" tab when logged in. The default Settings tab is "AI Providers Setup."

open_in_new

Note that, when you are on a non-settings page, the "Settings" tab will initially look like just a "gear" icon until you hover over it, when it expands to the full "Settings" tab with text label.

On the AI Provider Setup form, you'll pick the AI Provider that you want to connect GitClear with. After choosing your AI Provider, the form will update to include a link where you can retrieve your AI access key. Visit this URL, follow the applicable instructions, then paste your resulting AI API key into the form.

linkCursor API Stats

Cursor offers an excellent breadth of stats. Note that the "🟡" for "precise lines" reflects that Cursor offers an API to retrieve this, but it is limited to Enterprise customers.

To retrieve stats from Cursor, follow Cursor's help documentation to generate an Admin API key:

Format: key_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Start with "Dashboard" => "Settings":

open_in_new

Then scroll down the Settings page to "Advanced," where you can click the button for "New Admin API Key"

open_in_new

The key you copy can be pasted into GitClear's "Settings" => "AI Usage" tab.

linkClaude Code Analytics

GitClear has a plethora of options available for our Claude Code customers. Since, as of Q2 2026, Anthropic has lagged Cursor in delivering a rich set of usable per-committer data, most of the richest Claude Code ROI data emerges from installing GitClear's Claude telemetry script. At a high level, there are three integration options available to Claude Code subscribers:

Our usual recommendation: if you are the rare team with an Enterprise Team subscription, or with platform.claude.com as your primary entry point, add your Claude API key to GitClear's AI Provider Setup form.

Either way, use the Claude telemetry script if you want the most accurate AI ROI data. You can email hello@gitclear.com to get a link to download the script at no cost.

linkOption #1: Claude.ai with an Enterprise Team Plan

Claude.ai is where most users interact with Claude from. As of Q2 2026, Anthropic requires customers to purchase an Enterprise-level plan (minimum 20 seats, minimum one-year purchase) in order to receive an Analytics API key. If you have an Enterprise plan, click here to get instructions on retrieving your API key.

linkOption #2: Claude Console (platform.claude.com) Metered Plan

Claude Console (platform.claude.com) is where Admin API keys can be generated for any-sized team, if you elect to pay for your AI per-token. Check if you have stats visible on this page to confirm whether you are currently using Claude Console for code generation. Unlike Claude.ai Enterprise plans, Claude Console does not offer subscription plans - all billing is metered as pre-purchased credits.

linkOption #3: High Precision Code Line Telemetry

GitClear has built a proprietary Claude Telemetry implementation for any Claude Code team seeking detailed information about their Claude-authored code. Users of the telemetry solution will be able to directly assess the long-term outcomes for their code lines authored by Claude, with significantly higher confidence than is possible through a Usage-based API.

linkGithub Copilot Analytics & Measurement

If you log in via GitHub, no special setup is necessary on GitClear. You only need to ensure that your Github organization/enterprise enables "Copilot Metrics," as described in the following section.

If you log in via Bitbucket, Gitlab or Azure, you'll need to ensure Copilot Metrics are enabled, and either create a token, or log in to GitClear from your GitHub account.

linkEnsuring the enterprise account has Copilot Metrics "Enabled Everywhere"

To ensure "Copilot Usage Metrics" (the GitHub v2 AI API) is enabled,

open_in_new

Confirm that Copilot Metrics API and Copilot Usage Metrics are enabled

In some subscriptions, GitHub may not enable "Enterprise" access and may configure the Copilot Metrics API through Organization settings instead. In this case, you can view the Organizations you are able to configure at https://github.com/settings/organizations. Click "Settings" next to the organization that has Copilot seats, and choose "Copilot" => "Policies" in the left sidebar.

linkEnsuring that Organization(s) allow Copilot Usage Metrics

If you have newly enabled Copilot in your Enterprise, it's a good time to confirm that you have organizations also allow Copilot and Copilot Metrics to be retrieved:

open_in_new

Individual GitHub organizations need to have Copilot Usage Metrics enabled

Upon activating Copilot for the Enterprise, there is usually at least one Organization automatically enabled, but if you have multiple Organizations in your Enterprise, you likely need to visit those Organizations individually to enable Copilot is enabled.

After the Organization has Copilot enabled, you must then click on Copilot => Policies and confirm that "Copilot usage metrics" is enabled for this Organization. The process is identical to enabling Copilot Metrics for the Enterprise settings.

linkGetting Copilot Metrics with Bitbucket, Gitlab or Azure repos

You have two options to get Copilot usage metrics: you can either

The benefit of the latter is that it can be enabled immediately, but the drawback is that you either have to create a non-expiring token (which GitHub recommends against) or you have to undertake a new chore to go regenerate a token every X days.

On balance, we recommend logging in to GitHub and emailing us your two accounts to connect, but if you prefer to generate a token, read on.

linkGenerating a GitHub Personal Token to Access Copilot Metrics

A Github Personal Token (GitHub help page on the topic) can be used to generate AI Impact & Correlation Stats on GitClear.

First, on GitClear, visit your Entity's "Settings" => "AI Usage" tab.

open_in_new

Then, on GitHub:

open_in_new

Shown using a non-expiring token that can be deleted later by visiting GitHub Token Settings

After creating "Generate token," you should get a key of the length and prefix

ghp_XXXFGyxOsZpbdGIa4WV9KwGFIW6WIF1RXXXX

linkAugment Code Analytics

If you have an Enterprise subscription to Augment, you can connect your team's AI usage to GitClear. We recommend following Augment's latest instructions for retrieving your Augment token: the Augment documentation is available here. Augment's breadth of data offered is middle-of-the-road:

As of Q2 2026, the Augment setup looks like this:

open_in_new

Set up a Service Account, Generate an API token

Customers with successful Augment tokens retrieve a 64-character hex string to paste into GitClear's AI Provider Setup form. Augment supports per-developer AI usage stats, including counts of lines that have been changed by each LLM per day.

linkCodex by OpenAI Analytics

While OpenAI does offer a few Analytics APIs to their Enterprise customers, as of Q2 2026, none of these relate the count of "lines added" & "lines deleted" per committer. Since GitClear's AI ROI implementation relies upon mapping changed code lines to the specific LLM or human that changed them, this renders the Codex APIs not directly applicable for GitClear customers.

Instead of the limited OpenAI APIs, we have built a custom telemetry script that can be installed with Codex. Like it's Claude Code counterpart, the script works by deriving a one-way encrypted fingerprint of lines that are changed by OpenAI's LLMs. We pair the changed line fingerprints with metadata on which models are being used (and associated turn & token counts) in order to deliver sufficient granularity to ensure that decision makers can feel confident using the data to make decisions.

Please contact hello@gitclear.com if you would like access to our Codex telemetry script.

linkGemini Code Assist

GitClear supports integrating with Gemini Code Assist, an offering by Google that delivers Gemini LLMs within various IDEs. Thus far, Gemini Code Assist integration has a lighter variety of columns than other provides:

To retrieve your Gemini Code Assist credentials, create a service account in the Google Cloud Console with the "Monitoring Viewer" role, then generate and download a JSON key file. This JSON key file is what you will submit in GitClear's AI Provider Setup form.