Supported AI Tools and Provider Capabilities
Compare how LinearB connects to AI coding tools, what data each integration provides, and where the resulting insights appear. AI tool support is capability-specific. A tool can be visible through Gi…
Compare how LinearB connects to AI coding tools, what data each integration provides, and where the resulting insights appear.
AI tool support is capability-specific. A tool can be visible through Git-based attribution, provide first-party usage telemetry, connect to LinearB through MCP, or support a combination of these paths. An MCP connection does not mean that LinearB ingests that tool's usage telemetry.
How to read the matrices.
- Direct connector means LinearB collects data from the provider's API, exported reports, or OpenTelemetry.
- Git-based attribution means LinearB identifies AI involvement from signals such as commit co-authors, AI-authored pull requests, review comments, branch patterns, and AI rule files.
- MCP client means the tool can connect to the LinearB MCP server and query LinearB data. MCP is a separate capability from AI usage reporting.
AI analytics and telemetry capabilities
Use this matrix when choosing an integration for AI adoption, usage, workflow, and consumption reporting.
| AI tool | Collection path | Adoption and usage | Detailed activity | Tokens and estimated cost | Availability and cadence |
|---|---|---|---|---|---|
| GitHub Copilot | GitHub Copilot Usage API using a classic PAT, fine-grained PAT, or supported enterprise authentication. | Active users, adoption trends, suggestion acceptance, and AI code contribution. | Organization-level usage plus Git-based commit, branch, pull request, and review attribution. | Consumption (Estimated Cost). Availability of model- and token-level details depends on the data returned by GitHub. | Daily updates; up to 90 days of backfill. Some fine-grained and enterprise variants have limited availability. |
| Cursor | Cursor Admin API key plus Git-based attribution signals. | Active and engaged users, acceptance signals, and adoption reporting. | Accepted code lines, accepted tabs, command-line prompts, and correlated Git activity where available. | Consumption, token usage, and AI model reporting. | Supported Refresh and backfill limits are not specified in the current documentation. |
| Claude Code - Admin API | Anthropic Admin API key for organizations using an Anthropic API plan. | User-level daily usage, active users, adoption, and coding activity. | Sessions, commits, pull requests, accepted and rejected tool actions, accepted lines, and per-model activity. | Token breakdown and estimated cost by model. | Daily updates; up to 90 days of backfill. Not the setup path for AWS Bedrock. |
| Claude Code - Enterprise Analytics API | Anthropic Analytics API key with read:analytics for Claude Enterprise. |
Organization and user usage summaries, adoption, and activity reporting. | Summary-level analytics; not intended for detailed tool-action attribution. | Available when returned by the Claude API path; detail depends on Anthropic's analytics data. | Limited availability Three-day data delay; data begins January 1, 2026. |
| Claude Code - OpenTelemetry | LinearB ingest token and Claude Code OpenTelemetry configuration. | Claude usage and activity telemetry for AI Analytics and related reporting. | Direct telemetry with user attribution when user.email is configured. |
Consumption, tokens, models, skills, and tool calls. | Supported Recommended for Anthropic Teams, AWS Bedrock, and other third-party Claude providers. |
| Amazon Q Developer | Daily AWS user activity reports in S3, accessed through a read-only IAM role and Identity Store. | Usage and adoption signals in AI Analytics and related reporting. | Depends on the reports available in the configured S3 path and successful user resolution. | Not documented | Closed Beta Reports are generated daily in AWS. |
| Amazon Kiro | AWS usage reports in S3, accessed through an IAM role and Identity Store. | Usage and adoption metrics in AI Analytics. | Depends on the Kiro reports available in the configured S3 path. | Not documented | Open Beta |
| Codex | Codex Analytics API or OpenTelemetry, plus Git-based attribution signals. | Codex usage and activity reporting in AI Analytics. | Direct Codex usage data plus Git-based commit, branch, and pull request attribution where available. | Available when provided by the selected Codex connection type. | Depends on the selected connection type. |
| Gemini | Known-agent and Git-based detection where recognizable metadata or activity is present. | Agent activity may appear in AI adoption and workflow reporting when detected. | No direct Gemini telemetry connector is documented. | Not documented | Detection only in the current documentation. |
| Windsurf | No direct telemetry connector documented | Not documented as a first-party usage integration. | Git-based activity may be detected when known attribution signals are present. | Not documented | MCP client connectivity is supported separately. |
Use AI Analytics for current reporting. The standalone Copilot and Cursor Metrics dashboards and the legacy AI Tools Usage view were deprecated on April 2, 2026. AI Analytics provides team-scoped, repository-scoped, user-filterable, and pull request-level reporting.
Provider and connector variations
GitHub Copilot connection options
| Option | Use when | Status | Important notes |
|---|---|---|---|
| Classic PAT | Connecting standard GitHub organizations through the Copilot Usage API. | Supported | Requires Copilot Usage Metrics and the documented token scopes. Data updates daily with up to 90 days of backfill. |
| Fine-grained PAT | Organizations that require more granular permission control. | Limited availability | Requires both the token permission and an organization role with permission to view organization Copilot metrics. |
| Copilot Enterprise | Enterprise-wide reporting across teams and repositories. | Supported | Supports enterprise-wide reporting across organizations, teams, and repositories. |
| Enterprise Cloud with data residency | Enterprises hosted on a custom *.ghe.com subdomain. |
Supported | Routes metrics requests to the enterprise-specific API base URL. Standard GitHub-hosted enterprises use the standard setup. |
Claude Code connection options
LinearB supports one Claude connector at a time. Choose the path that matches your Claude environment and reporting needs.
| Option | Recommended environment | Reporting depth | Timing |
|---|---|---|---|
| Admin API | Anthropic API organizations. | Richest current path: user-level usage, sessions, Git activity, tool actions, accepted lines, models, tokens, and estimated cost. | Daily; up to 90 days of backfill. |
| Enterprise Analytics API | Claude Enterprise organizations. | Organization and user usage summaries focused on adoption, not detailed event-level telemetry. | Three-day delay; available analytics data begins January 1, 2026. |
| OpenTelemetry | Anthropic Teams, AWS Bedrock, and other third-party Claude providers. | Direct usage and activity telemetry. Per-user attribution requires a unique user.email value. |
Begins after the configuration is deployed; historical backfill is not documented. |
Codex connection options
LinearB supports multiple Codex connection methods. Choose the path that matches how your organization collects and manages Codex usage data.
| Option | Recommended environment | Reporting depth | Timing |
|---|---|---|---|
| Analytics API | Organizations connecting Codex directly to LinearB through the Codex Analytics API. | Codex usage and activity data available through the Analytics API. | Refresh and historical backfill details are not yet documented. |
| OpenTelemetry | Organizations sending Codex telemetry directly to LinearB using OpenTelemetry. | Direct Codex usage and activity telemetry. | Begins after OpenTelemetry is configured; historical backfill details are not yet documented. |
LinearB MCP client compatibility
The LinearB MCP server lets supported clients query LinearB metrics, repositories, contributors, incidents, teams, pull requests, project-management entities, and other engineering data using natural language.
| Client | MCP support | Authentication | Configuration path |
|---|---|---|---|
| Claude Desktop | Supported | OAuth recommended; API key available where required. | Custom connector in Claude Desktop. |
| Claude Code | Supported | Browser-based authentication after CLI registration; API key is also documented. | claude mcp add |
| Cursor | Supported | OAuth recommended; API key alternative. | Cursor MCP server configuration. |
| Codex | Supported | Browser-based OAuth login. |
codex mcp add and codex mcp login
|
| Windsurf | Supported | Native OAuth. | Windsurf MCP server configuration. |
| GitHub Copilot | Not documented | - | - |
| Gemini | Not documented | - | - |
| Amazon Q Developer / Kiro | Not documented | - | - |
The MCP server is documented as a Beta feature. Verify important metrics in the LinearB UI before using MCP-generated results for critical reporting or decisions. Cline, Continue.dev, VS Code, and Zed are also documented MCP clients.
Capabilities that are independent of the external AI provider
LinearB AI Review and AI Description
LinearB AI Review and AI Description are LinearB-managed automations. They do not require the organization to connect GitHub Copilot, Cursor, Claude Code, Amazon Q Developer, or Amazon Kiro for usage reporting. AI Review is powered by Claude 4.5 Sonnet and supports mainstream programming languages; test less common languages with a representative pull request.
AI Services
AI Services is the organization-wide switch for LinearB AI processing and managed automations. When AI Services is disabled, AI Review, AI Description, AI-generated summaries, and other LinearB AI processing stop. External GitHub Copilot, Cursor, and Claude usage data can continue to report because those integrations rely on provider data rather than LinearB AI processing.
AI attribution
Direct provider telemetry is not required for every attribution use case. LinearB can also identify AI involvement through co-author metadata, AI-authored pull requests, review comments, branch prefixes, and repository rule files. Co-author metadata is the strongest direct Git signal; tool usage and timing correlation can improve coverage when direct metadata is missing.
Provider data differs. Metrics such as suggestion acceptance, sessions, tool calls, model usage, token consumption, and estimated cost are available only when the selected provider and connector expose the required data. A blank or unavailable metric does not necessarily mean the AI tool was not used.
How did we do?
Connect Kiro to LinearB
Troubleshoot GitHub Copilot integrations
