Integrations

Last updated: Edit on GitHub

Plan Availability

IntegrationStarterProEnterprise
GitHubYesYesYes
SlackYesYesYes
MLflowNoYesYes
Weights & BiasesNoYesYes
JiraNoYesYes

All integrations are configured under Settings > Integrations in your workspace. You must have Admin or Owner role to add or remove integrations.

GitHub

The GitHub integration automatically syncs your AI system’s compliance status when model-related code is pushed to a connected repository. It is most useful for teams who version their model code, training scripts, and configuration files in Git.

What it does

  • Detects pushes to a configured branch (e.g., main or release/*) and creates a new system event in the Aikraft monitoring timeline
  • Reads model-card.md or aikraft.yaml files from the repository root, if present, and uses them to auto-populate documentation fields
  • Posts a compliance status check to pull requests, showing the current risk tier and documentation score (useful for blocking deployments when documentation is incomplete)

Setup

  1. Go to Settings > Integrations > GitHub.
  2. Click Connect to GitHub. You will be redirected to GitHub to authorise the Aikraft GitHub App.
  3. Select the repositories you want to grant access to. You can select individual repositories or grant access to all repositories in an organisation.
  4. Return to Aikraft and map each connected repository to an AI system in your workspace using the dropdown.
  5. Select the branch to monitor (default: main).
  6. Optionally enable PR Compliance Checks to post a status check on every pull request.
  7. Click Save.

The first sync runs immediately after saving. Subsequent syncs are triggered by push events via webhook.

Slack

The Slack integration sends compliance alerts and monitoring notifications directly to a Slack channel of your choice. This is the recommended way to ensure your team acts on critical alerts promptly.

What it does

  • Posts alerts when monitoring thresholds are breached (data drift, prediction drift, incident logged)
  • Notifies the team when documentation score drops below a configured threshold
  • Sends a weekly compliance digest summarising the health of all monitored systems
  • Mentions specific users or groups when an alert requires action (configurable per alert type)

Setup

  1. Go to Settings > Integrations > Slack.
  2. Click Connect to Slack. You will be redirected to Slack to authorise the Aikraft app for your workspace.
  3. Select the default notification channel (e.g., #ai-compliance).
  4. Optionally configure per-system channel overrides — useful if different teams own different systems.
  5. Click Save.

To configure which alert types are sent to Slack, go to Settings > Notifications and toggle the Slack switch for each alert type. You can also set a minimum severity level so that only Warning and Critical alerts are posted, keeping the channel signal-to-noise ratio high.

MLflow

The MLflow integration connects Aikraft to your MLflow Model Registry, pulling model version metadata into the monitoring timeline and triggering re-documentation when new model versions are registered.

What it does

  • Polls the MLflow Model Registry hourly for new model versions
  • Records model version, run ID, parent experiment, and key metrics (accuracy, F1, AUC) in the Aikraft monitoring timeline
  • Triggers a re-documentation prompt when a model version tagged Production or Staging is detected
  • Pulls metric history charts from MLflow and embeds them in the Annex IV Section 5 (Performance Metrics) documentation draft

Setup

  1. Go to Settings > Integrations > MLflow.
  2. Enter your MLflow Tracking Server URI (e.g., https://mlflow.yourcompany.com or a Databricks workspace URL).
  3. Choose authentication method: Personal Access Token or Service Account Token (recommended for production).
  4. Enter the token and click Test Connection. Aikraft will verify it can reach the server and read experiment data.
  5. Select the registered model name to link to your Aikraft system.
  6. Click Save.

If your MLflow server is on a private network, you will need to allow inbound connections from Aikraft’s static IP range. Contact aikrafteu@proton.me for the current IP list.

Weights & Biases

The Weights & Biases (W&B) integration pulls experiment run data and model performance metrics into Aikraft, providing a richer picture of how your model’s behaviour evolves over time.

What it does

  • Syncs run metadata and logged metrics from a configured W&B project into the monitoring timeline
  • Automatically populates Annex IV Section 5 (Performance Metrics) with metrics from your production-tagged runs
  • Detects significant metric regressions between runs and surfaces them as monitoring alerts

Setup

  1. Go to Settings > Integrations > Weights & Biases.
  2. Enter your W&B API Key (found at wandb.ai/settings).
  3. Enter the Entity (your W&B username or team name) and Project name.
  4. Choose which runs to track:
    • All runs: every run in the project is synced
    • Production-tagged runs only: only runs with the tag production are synced (recommended)
  5. Click Test Connection, then Save.

Metrics are synced every hour. The first sync pulls the last 30 days of run history.

Jira

The Jira integration creates Jira tickets automatically when compliance tasks are generated in Aikraft — for example, when a documentation section is flagged as incomplete, when a monitoring alert requires remediation, or when a re-classification is needed.

What it does

  • Creates a Jira issue for each compliance action item, with the Aikraft system name, issue description, severity, and a direct link back to the relevant page in Aikraft
  • Updates the Jira issue status when the corresponding action is completed in Aikraft (e.g., documentation score reaches 95+)
  • Assigns issues to specific Jira users based on a mapping you configure (e.g., all documentation tasks to the compliance officer, all drift alerts to the ML engineer)

Setup

  1. Go to Settings > Integrations > Jira.
  2. Enter your Jira instance URL (e.g., https://yourcompany.atlassian.net).
  3. Click Connect to Jira. You will be redirected to Atlassian to authorise the Aikraft app.
  4. Select the Jira project where issues should be created.
  5. Map Aikraft alert severity to Jira issue priority:
    • Critical → Highest
    • Warning → Medium
    • Info → Low
  6. Optionally configure user mappings under Assignee Rules.
  7. Click Save.

Jira issues created by Aikraft include the label aikraft-compliance for easy filtering. You can search for all open Aikraft issues in Jira using the JQL query:

labels = "aikraft-compliance" AND resolution = Unresolved ORDER BY priority DESC

To close the loop, mark an Aikraft alert as resolved in the Aikraft UI — the linked Jira issue will be automatically transitioned to Done (requires that your Jira project’s workflow includes a Done status reachable from the current status).