How-to · FaultAtlas
Configure FaultAtlas with an assistant
A safe, verifiable path for using an assistant while configuring shipped FaultAtlas integrations.
Use an assistant to explain versioned FaultAtlas documentation or draft configuration, but keep every credential and environment-specific value outside the prompt.
Choose the integration
- Use the Node.js Agent for server exceptions and messages.
- Use the React and Browser Agents for browser failures, navigation breadcrumbs and Web Vitals.
- Use the Next.js or Vite development adapter only for the documented local debug workflow.
- Use the project-scoped Agent API or MCP server only when the documented permission model fits the task.
Keep secrets local
Never paste credentials into third-party prompts. Keep ingest keys, session tokens, API tokens, OAuth secrets and database URLs in the local or server environment that consumes them. Give an assistant placeholders such as FAULTATLAS_INGEST_KEY, never real values.
Ask for a bounded change
Provide the assistant with the relevant versioned guide URL, framework and runtime. Ask it to change only the documented integration point and to preserve the existing error handling and redaction configuration.
Verification
- Review the proposed diff before running it.
- Confirm required environment variables are present without printing their values.
- Run the package tests and typecheck documented for the integration.
- Send a synthetic, non-sensitive failure and confirm it appears in the intended Project.
- Remove the synthetic event if your retention policy requires it.
If verification fails, stop. Keep the failure type and source location, remove any secret values, and compare the implementation with the matching versioned guide. Do not weaken redaction, authentication or TLS to make setup pass.