AI Coding Agents

Pasting a stack trace into Cursor by hand is a start. It's not the same as giving an agent the full picture of what happened. LogRocket's AI coding agents do the latter — they turn any issue into a fix-ready prompt, pre-loaded with the session activity, network requests, console output, and component context an engineer would normally gather by hand.

There are two ways to use them:

  • Open in a coding tool — open the issue in your local AI coding tool (Cursor, Claude CLI, Codex, or Windsurf) so you can debug and fix it interactively. Learn more →
  • Dispatch a cloud agent — send the issue to a Cursor cloud agent that opens a pull request without you leaving LogRocket. Dispatch manually from any issue, or enable auto-dispatch to have agents launch automatically when Galileo detects a new severe issue. Learn more →

Which one should I use?

Open in a coding toolDispatch a cloud agent (manual)Auto-dispatch
Where it runsYour machineCursor's cloudCursor's cloud
Supported toolsCursor, Claude CLI, Codex, WindsurfCursor (more coming)Cursor (more coming)
TriggerYou open the issue in your toolYou click Send to Cursor Cloud AgentGalileo detects a new severe issue
SetupNone requiredProject-wide API key in Settings or personal override API keyGalileo + project-wide API key + Issues Settings toggle
API key usedN/AProject-wide or personal overrideProject-wide only
Best forHands-on debugging, exploration, any code change you want to driveRoutine fixes you want to delegate, working in parallel on multiple issuesHands-off fixes on the most critical new issues
OutputThe issue opens in your tool with a prepared promptA PR (or work-in-progress branch) from the cloud agentA PR (or work-in-progress branch) from the cloud agent

What makes the prompt different

Galileo AI generates the prompt — and it's not a one-line stack trace. It includes a link to a debug package the agent can unpack and read on its own.

The debug package contains the context an engineer would normally have to gather by hand:

  • The error and where it occurred — the exact exception, file, and line
  • The user actions that led to the error — the relevant session activity in sequence
  • Network requests and responses — what was in-flight around the time of the error
  • Console output — logs and warnings leading up to the failure
  • Component, route, and environment information — the state of the app at the moment it broke

Say a user hits a checkout error at 2am. Galileo packages the stack trace, the two failed API calls that preceded it, the console warnings from the payment component, and the route they were on. When you open the issue in Cursor, the agent already understands the failure path — not just what threw, but what caused it to throw.

That's the difference: a stack trace tells the agent what failed. The debug package tells it why.

Permissions and access

  • Open in a coding tool is available to anyone who can view an issue. No integration setup required.
  • Dispatch a cloud agent requires a Cursor API key configured by an admin under Settings → Integrations → Coding Tools → Cursor for project-wide access for your whole team. Individual users can set their own Cursor API key via User Settings; a personal key overrides the project-wide key for manual dispatch only.
  • Auto-dispatch requires Galileo AI enabled, a project-wide Cursor API key, and the auto-dispatch toggle in Settings → Issues Settings. It always uses the project-wide key — personal overrides do not apply. Learn more →

Get started