← Back to all sparks
B

Bugsnag

INFRA · APIS
Velocity1.7

Bugsnag is wiring AI agents directly into the debug loop via MCP.

ai observabilitymcp integrationerror monitoringmobile sdksperformance monitoringdata access api
Current state
Bugsnag's monthly cadence is locked onto AI-workflow integration as the central theme. The MCP server has grown from a query bridge into something agents can act through—Fix-with-MCP shipped as a first-class resolution flow in December, then picked up Jira-linking and snooze tools, and now supports OAuth for self-hosted. Around that core, mobile and game observability keep expanding (Flutter perf, Unreal 5.7, Vega OS, App Hang detection, FPS telemetry), and the dashboard is gaining Advanced Search, Performance Score, and Correlated Events for richer signal shaping.
Where it's heading
The product is converging toward observability data that AI clients can both read and act on. Every recent release ties back to that loop: SDK additions expose more controllable error metadata, the Data Access API keeps gaining surface (commenting, project-by-API-key lookup), and MCP gets new verbs and auth options. Non-AI work like Correlated Events and HTTP attribute tracking feeds the same agenda by producing the kind of structured signal an agent—or a human—can pivot on.
Prediction
Expect deeper Fix-with-MCP automation next (auto-triage, suggested fixes pushed into PRs) and a richer Data Access API for AI clients, likely paired with another platform addition on the mobile or device side to keep the surface-area story moving.

Recent moves

  1. 9d ago

    Correlated Events, MCP OAuth, and Flutter perf metrics in April

    Correlated Events groups errors sharing a user, device, session, or trace—filling in a traceability gap that nudges Bugsnag toward APM-style debugging. Self-hosted MCP gaining OAuth and Flutter getting CPU/memory/FPS metrics fit the same surface-expansion arc.

    View source ↗
  2. 17d ago

    Android App Hang detection and Data Access API comments in March

    Android App Hang detection lands, the Data Access API gains commenting, and HTTP-request error reporting reaches Android and Cocoa SDKs. Each piece feeds the same loop: more structured signal for the AI/MCP flows Bugsnag has been building.

    View source ↗
  3. 2mo ago

    MCP gains Jira linking and snooze; source-maps package retired

    MCP picks up snooze and Jira-linking tools, the Data Access API gets trace/session/device IDs, and the old @bugsnag/source-maps package is retired in favor of the BugSnag CLI. Continues to fill out the AI-actionable surface around errors.

    View source ↗
  4. 3mo ago

    SDK delivery controls and isFatal flag in January release

    Android and Cocoa SDKs add a deliveryStrategy field and isFatal flag, giving developers finer control over when events are sent and whether handled exceptions are escalated. React Native performance gains native app-start coverage. Incremental but useful SDK ergonomics work.

    View source ↗
  5. 4mo ago

    Fix with MCP launches: AI-powered error resolution

    ⚡ SPARK

    Fix-with-MCP debuts as a first-class AI-resolution action with its own project settings page, and Performance Score joins Stability Score as a standardized health metric. This is the inflection point the trajectory above points back to—the release where AI moves from auxiliary to default.

    View source ↗
  6. 5mo ago

    Separate symbol-upload API keys for tighter CI credentials

    Separates the symbol/mapping upload API key from the main project key—small but useful for teams locking down CI credentials. SmartBear ID logins reach BugSnag for Enterprise. Quiet release in the run-up to the December AI push.

    View source ↗