> For the complete documentation index, see [llms.txt](https://docs.burpgpt.app/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.burpgpt.app/help-and-faq/troubleshooting.md).

# Troubleshooting

Resolve common installation, provider, scanning, and AI Chat problems.

Inspect provider requests and responses in Burp's `Logger` when diagnosing model connection problems.

Under `Extensions` → `Installed`, select BurpGPT Pro and check `Errors` for startup and operation failures. Check `Output` for messages prefixed with `Warning:` and follow any suggested action.

## The extension does not load

* Use Burp Suite Professional 2026.7 or later.
* If you launch Burp with an external Java runtime, use Java 21 or later.
* Check the extension error output. If the JAR is damaged, download a fresh copy.

See [Installation](/getting-started/installation.md).

## The extension loads, but features are not visible

Open `BurpGPT Pro` → `Licence` and check the displayed status. See [Activate your licence](/getting-started/activate-your-licence.md) or [Why did licence activation fail?](/help-and-faq/faq/what-are-the-possible-causes-for-an-unsuccessful-activation-of-my-burpgpt-pro.md).

## Provider settings reset after upgrading

Version 3.1 resets provider and scanner settings from earlier releases. Re-enter the provider configuration, run `Test request`, and review your scanner preferences. Saved licences and prompt libraries are retained.

## Settings cannot be saved

Provider and scanner settings are saved automatically as you edit them. If a save fails, the page displays an inline error and the extension records the failure in `Errors`. The failed edit has not been saved, even if it remains visible in the field. Check the error output before continuing.

## `Test request` fails

Inspect the provider response in `Logger`, then check:

* The provider matches the endpoint, and `Base URL` is correct.
* `API key` is present when required.
* `Model` is an exact identifier available to your account or model server.
* You pressed Enter after typing a custom model identifier; the dropdown's examples do not confirm availability.
* `Request parameters` contains valid JSON and [supported fields](/getting-started/supported-model-providers.md#request-parameters) with the correct types.
* The model supports the parameters you supplied. Try `{}` to remove optional parameters.
* Your provider has sufficient quota, credits, or local resources.
* `Read timeout` is long enough for the model to respond.

For HTTPS endpoints, certificate verification must succeed. If an HTTPS connection fails, check the certificate, hostname, and trust configuration.

See [Test your provider](/model-setup/test-and-validate-model-provider-settings.md).

## A local provider does not connect

Confirm that the model server is running and reachable from the computer running Burp. If it uses another host or port, update `Base URL`.

For Ollama, use `http://localhost:11434` for a standard local installation and confirm your model is available with `ollama ls`.

For OpenAI-compatible APIs, use the API base URL, typically ending in `/v1`, rather than the full `/chat/completions` URL.

## A test succeeds, but a scan or chat request fails

The actual prompt may be larger than the test request. Inspect the provider response for a context-limit error or timeout. Use narrower [placeholders](/analysis-and-prompts/use-placeholder-reference.md), reduce the included traffic, or increase the timeout as appropriate.

`max_tokens` limits generated output; it does not reduce the input prompt.

## Scans do not produce results

* Confirm that `Test request` succeeds.
* Review the role and prompt in `AI scanner`.
* For automatic analysis, enable `Scan: On` and start an active scan in Burp. The toggle does not scan Proxy history on its own.
* For targeted analysis, use `Extensions` → `BurpGPT Pro` → `Scan with AI (results in Target → Site map)`.
* Look for **GPT-generated insights** issues in `Target` → `Site map`.

See [Analyse HTTP traffic](/analysis-and-prompts/analyse-http-traffic.md).

## AI Chat does not include the expected traffic

Select `Request`, `Response`, or both before sending your message. A response is only available when one exists for the selected request.

Included traffic toggles are deselected when you send a message, even if the model request later fails. Select them again to attach the current traffic to another message. Previously sent traffic can still be included in conversation history: unticking Request or Response does not remove it. Selecting another request in the same editor does not clear the conversation either; use `Clear chat` before a separate investigation.

Each model request includes up to 15 previous completed exchanges plus your new message. Older exchanges remain visible but are no longer sent. See [conversation context](/analysis-and-prompts/use-ai-chat.md#conversation-context).

## A prompt import fails or replaces an entry

Import a JSON array in the [prompt library format](/analysis-and-prompts/use-the-prompt-library.md#import-prompts). Each entry needs non-blank `author`, `category`, and `text` fields.

Imports update existing entries with the same ID. Different IDs remain separate, even if the prompt text matches. Export a backup before importing or resetting the library.