Intel Engine vs Explain Mode
CrashCatch has two analysis layers that work independently of each other.
- Intel Engine runs automatically on every crash dump. It uses heuristics pattern matching against exception codes, stack shapes, and register values to classify the crash and flag indicators. No internet connection or API key required.
- Explain Mode sends a structured summary of the crash to the Claude AI and returns a plain-English explanation. It requires an Anthropic API key and an internet connection.
For most crashes, Intel will give you a fast, accurate classification. Explain Mode is most useful when you want a narrative explanation, or when working with a crash type you haven't seen before.
Setting up Explain Mode
Step 1 Get an Anthropic API key
Explain Mode uses the Anthropic Claude API. You need your own API key.
- Go to console.anthropic.com and sign in or create an account
- Navigate to API Keys in the left sidebar
- Click Create Key, give it a name (e.g. "CrashCatch"), and copy the key
Note that Anthropic API usage is billed per token. Explain Mode uses a compact model and keeps requests small typical usage costs a fraction of a cent per analysis.
Step 2 — Enter the key in CrashCatch
- Open CrashCatch Analyze
- Go to Tools > Preferences
- Paste your API key into the Anthropic API Key field
- Click Save
Your key is stored locally in the app's storage it is never sent to CrashCatch Labs servers.
Running an explanation
With a crash dump open and an API key configured:
- Click the Explain tab, or
- Go to View > Explain Mode
CrashCatch will send a structured summary exception code, crashing thread stack, Intel findings, and key register values to Claude and display the response when it arrives. This typically takes a few seconds.
Auto-explain on open
If you want an explanation to run automatically every time you open a file, enable Auto-explain on open in Tools > Preferences. This is most useful if you're working through a batch of crash dumps and want the explanation ready as soon as the file loads.
Understanding the output
Explain Mode returns three structured sections:
- Root cause a concise sentence identifying what caused the crash. For example: "A null pointer dereference in the mesh rendering pipeline, most likely caused by accessing a component that has already been garbage collected."
- Evidence the specific signals from the dump that support the conclusion: exception code, fault address, top stack frames, and anything unusual in the register state.
- Suggested fix concrete next steps. These might include adding a null check, reviewing the call site in a specific frame, or checking for a race condition in a particular subsystem.