gitStream - Start Here

gitStream is LinearB’s automation engine for pull requests. It runs in your Git provider and lets you:

• Automate PR policies (size, ownership, risk, approvals).
• Route reviews to the right people based on expertise, files, or teams.
• Apply AI to descriptions and reviews to reduce manual toil.
• Enforce consistency while keeping PR noise as low as possible.

Instead of every team reinventing PR rules in different repos, gitStream centralizes your policies as code and applies them consistently across your organization.

AI adoption is accelerating—but most engineering leaders still struggle to measure its true impact. LinearB helps you quantify how AI changes your workflow, code quality, delivery speed, and downstream stability.

Key questions LinearB answers:

• Delivery speed
  • Is AI helping my teams ship faster?
  • How much faster—and where?
  • If not faster, what’s blocking improvements?
• Code volume & merge efficiency
  • How much AI-generated code is being created?
  • How much of it actually gets merged?
  • How does AI-assisted code behave compared to non-AI code?
• Stability & incidents
  • Which AI-assisted PRs reach production?
  • Which of them introduce incidents or rollbacks?
  • What’s the downstream impact on MTTR and CFR?
• Code quality & maintainability
  • Is AI introducing:
    • Bugs
    • Security vulnerabilities
    • Maintainability issues
    • Performance regressions
    • Readability or scope problems
  • Do AI-generated changes result in higher churn?
• Team impact
  • Which teams leverage AI effectively?
  • Which teams see bottlenecks, slowdowns, or increased rework?
  • How does AI adoption affect reviewer workload?

How LinearB & gitStream enable this:

• Tag PRs influenced by AI tools or AI-generated code.
• Measure how AI-influenced PRs move through the funnel (Open → Merge → Release → No Rollback).
• Track performance differences between AI and non-AI changes.
• Identify which AI-assisted PRs contribute to production issues.

With LinearB, AI shifts from guesswork to accountability—giving you clear visibility into the value and risks introduced by AI-driven coding.

At a high level, gitStream is designed to respect the same security and privacy principles as the rest of LinearB:

• Read-only access: gitStream and LinearB read repository metadata and code only as needed to evaluate rules and AI prompts. They do not modify your code.
• Least privilege: Installation scopes are limited to what is required for PR automation (for example: reading PRs, comments, reviews, and status checks).
• Config as code: Your policies live in configuration files (YAML) in your repos or central configuration, giving you full visibility and change control.
• AI usage transparency: Where AI is used (descriptions, review suggestions, summaries), gitStream clearly identifies AI-assisted actions.

For formal details, please refer to your LinearB's security portal. For access, please contact support

Time: 2–5 minutes  |  Difficulty: ⭐ Easy

High-level flow:

1. Go to LinearB → Automations → gitStream.
2. Select GitHub Cloud and click Install App.
3. Choose your GitHub organization.
4. Install for all repositories (recommended) or a subset.
5. Return to LinearB and confirm repositories appear under gitStream.

Verification: Open a test PR and confirm gitStream automations fire once YAML or Managed Mode rules are enabled.

Full guide: GitStream for GitHub Cloud — Installation Guide

Time: 5–10 minutes  |  Difficulty: ⭐⭐ Medium

High-level flow:

1. Create a Custom GitHub App in your Enterprise instance.
2. Configure required permissions (PRs, comments, checks, labels).
3. Set webhooks to the endpoint provided by LinearB.
4. Generate a private key and copy the App ID / secrets into LinearB.
5. Install the app on the target organization(s).

Full guide: GitStream for GitHub Enterprise Server — Installation Guide

Time: 10–15 minutes  |  Difficulty: ⭐⭐ Medium

High-level flow:

1. Create a Personal Access Token with api scope.
2. In LinearB, go to Automations → gitStream and select GitLab.
3. Enter the token and GitLab URL (for self-managed).
4. Select which projects/repositories gitStream should monitor.
5. Configure webhooks / CI according to the detailed GitLab guide.

Full guide: GitStream for GitLab — Installation Guide

Time: 10–15 minutes  |  Difficulty: ⭐⭐ Medium

High-level flow:

1. Create a Bitbucket OAuth Consumer.
2. Copy the Client ID and Secret into LinearB under gitStream.
3. Authorize LinearB to access the workspace.
4. Add webhooks for the repositories you want to automate.

Full guide: GitStream for Bitbucket Cloud — Installation Guide

Time: 1–3 minutes  |  Difficulty: ⭐ Very Easy

Where to configure:
Go to Company Settings → AI Tools.

1. Enable Managed Mode (Essentials only)

1. Navigate to Company Settings → AI Tools.
2. Select Managed Mode as the execution mode.
3. Toggle on the automations you want (e.g., AI Review, AI Description, Co-author labels, ETR labels, Dependabot approvals).

Managed Mode uses LinearB’s managed runner. No YAML or CI changes are required.

Full Essentials setup guide: Install gitStream entirely from the LinearB Platform

2. (Optional) Use Self-Managed Mode with YAML

If you prefer full control, you can switch to Self-Managed Mode and configure gitStream via YAML files in your repositories (.cm). Use the starter YAML below as a low-noise baseline.

Once gitStream is installed, you can layer in more automations to reduce toil and improve consistency:

• Reviewer assignment based on expertise (paths, services, or ownership).
• Fast-track PRs that are small, low-risk, or docs-only.
• Enforce commit or branch naming conventions.
• Standardize PR templates and required metadata.
• Deeper AI-based review rules for security, performance, or maintainability.

Recommended next reads:

Tip: Start with 1–2 high-value automations, validate the impact with a pilot team, then roll out to more repos.

• Confirm gitStream is installed and connected for your Git provider (see provider section above).
• For Self-Managed Mode (Essentials or Enterprise plan), verify:
  • Your YAML files are in the default branch under .cm/.
  • The on: triggers (e.g., pr_created, commit, pr_ready_for_review) match the events you expect.
  • There is no if: condition that always evaluates to false (for example, restricting to a branch that is not used).
• For Managed Mode (Essentials only), check that:
  • Managed Mode is selected under Company Settings → AI Tools.
  • The specific automations (AI Review, AI Description, etc.) are toggled ON.
• Confirm webhooks or app events are successfully reaching LinearB (no 4xx/5xx errors in your Git provider’s webhook logs).

• Make sure the describe-changes@v1 action exists in your YAML and is under an automation whose triggers run on PR creation or commits.
• Verify concat_mode: append (or your chosen mode) is configured correctly.
• Check that the PR is not marked as draft if your if: clause excludes drafts.
• Ensure the gitStream app/token has permission to update PRs in your Git provider.

• Confirm the automation uses add-label@v1 or add-labels@v1 with valid label names.
• Check that you are on a provider that supports labels:
  • GitHub / GitLab: labels supported.
  • Bitbucket Cloud: labels on PRs are not supported; label-based actions will not apply.
• Verify the app/token has permission to manage labels.
• Check when: conditions on the label actions (for example, unresolved threads, missing Jira) are actually true for the PR being tested.

• Confirm the YAML uses add-reviewers@v1 and that who.experts (or your reviewer list) is defined in the config section.
• Ensure gitStream has permission to request reviewers in your Git provider.
• Check that the PR author is not listed as a reviewer (some providers ignore self-assignment).
• Verify the automation’s if: conditions are met (for example, not a bot, not a draft).

• Confirm your YAML has a single add-comment@v1 action and that no other automations are also posting summary comments.
• Make sure you’re not using conflicting rules that run on overlapping triggers (e.g., multiple automations with pr_created + commit both adding comments).
• If you recently changed the comment template, test on a new PR to ensure old comments are not being confused with the new format.

If you’ve confirmed installation, webhooks, and YAML but gitStream still isn’t behaving as expected, please contact LinearB Support with:

• Visit our extneral gitStream documentation