Skip to content
For the complete documentation index, see llms.txt. This page is also available as Markdown.

Which Build?

Custom AI Agent ships as two builds from one repository, the BApp Store build and the GitHub full build: pick one and load only that one.

The two builds

BApp Store build
GitHub full build

Gradle task

./gradlew bappJar

./gradlew fullJar

Output file

build/libs/custom-ai-agent-bapp.jar

build/libs/Custom-AI-Agent-full-<version>.jar

Release asset

Custom-AI-Agent-bapp-<version>.jar and Custom-AI-Agent-bapp-<version>.jar.sha256

Custom-AI-Agent-full-<version>.jar and Custom-AI-Agent-full-<version>.jar.sha256

Where to get it

GitHub Releases today; the BApp Store once PortSwigger lists it (see below)

MCP tools

8 extension-native tools: status, issue_create, ai_analyze, ai_passive_scan, ai_findings_recent, redact_preview, ai_audit_query, ai_backends_list

All 59: the same 8 plus 51 generic Montoya API wrappers

AI backends

7: Burp AI, Anthropic, Ollama, LM Studio, NVIDIA NIM, Perplexity, Generic OpenAI-compatible

12: the same 7 plus Gemini CLI, Claude CLI, Codex CLI, OpenCode CLI, Copilot CLI

External MCP servers

Not included

Included: External MCP Servers and their ext:<server>:<tool> tools

Passive-to-active

Suggested targets, queued by hand ("Queue N suggested targets" in the AI Active Scanner Queue dialog)

The same, plus the opt-in "Auto-queue passive findings for active testing", off by default

Burp "Use AI"

Needed by every AI backend, Burp AI included, and by MCP ai_analyze and ai_passive_scan

Needed by Burp AI and by MCP ai_analyze and ai_passive_scan; the HTTP and CLI backends do not depend on it

Health checks

On demand only: Save, a backend change, a click on the status pill, Test connection or the first chat message. The pill reads "Not checked ยท click to test" until then

A startup check for every backend, then every 30 s for local backends (Burp AI, the CLI backends, HTTP servers on a loopback address); remote providers again after a settings or backend change and on a pill click

Drop-in backend JARs

Not supported: the build never creates, lists or loads ~/.burp-ai-agent/backends

Loaded from ~/.burp-ai-agent/backends (Adding a Backend)

AI scanners at load

OFF at every load; switching one on lasts until the extension unloads

The same

Name in Burp

Custom AI Agent; the Output tab banner reads Custom AI Agent <version> (BApp Store build)

Custom AI Agent; the Output tab banner reads Custom AI Agent <version> (GitHub full build)

The Settings > Help caption names the running build the same way. In Settings > MCP Server, the native tools are listed under "Extension-native tools โ€” available in both the BApp Store and the full build." and the generic tools under "Generic Montoya API wrappers โ€” available in the full build only."

Burp edition requirements

The BApp Store build runs on Burp Community and Burp Professional; it sends AI requests only while Burp's "Use AI" setting is enabled for the extension, and with it off it loads and refuses every AI request.

The BApp Store build checks Burp's "Use AI" setting through api.ai().isEnabled() at each AI request. While it is off, the extension still loads and its MCP server, native tools and privacy settings keep working, but every AI request is refused and the status pill shows "AI disabled in Burp". Which Burp editions can turn "Use AI" on is decided by Burp, not by this extension, and it was not measured on Burp Community. For local or CLI AI that does not depend on "Use AI", use the GitHub full build: its HTTP and CLI backends do not depend on that setting, while Burp AI and the MCP tools ai_analyze and ai_passive_scan still do.

What each build does with "Use AI" on and off. This is the gate the code enforces; it is not a statement about any Burp edition:

"Use AI" for the extension
BApp Store build
GitHub full build

On

AI requests go to the selected backend

AI requests go to the selected backend

Off

Loads. The MCP server, the native tools and the privacy settings keep working. Every AI request is refused: the chat shows the refusal below, the status pill reads "AI disabled in Burp" and the scanner panel reads "Paused: Use AI is off"

Loads. Burp AI is refused; the HTTP and CLI backends keep working. MCP ai_analyze and ai_passive_scan return a tool error with the reason

The BApp Store build's refusal in the chat reads: "This build sends AI requests only while Burp's 'Use AI' setting is enabled. Turn it on in Burp's settings to use any AI backend."

Where "Use AI" lives. PortSwigger documents "Use AI" as a per-extension checkbox, off by default: Extensions > Installed > Custom AI Agent > Use AI.

BApp Store or GitHub Releases

Custom AI Agent is not listed in the BApp Store yet: the BApp Store build is going through PortSwigger's review in extension-portal#231.

From v1.1.0, every GitHub release attaches both JARs, Custom-AI-Agent-bapp-<version>.jar and Custom-AI-Agent-full-<version>.jar, each with a .sha256 file in sha256sum format, plus one bom.json (CycloneDX SBOM). Check a download with sha256sum -c <asset>.sha256 in the folder that holds both files. The checksum is published on the same release as the JAR, so it shows the file is complete, not who built it. bom.json lists the full build's runtime dependencies, which include every dependency of the BApp Store build.

Load only one build

Both builds share Burp's saved settings and encrypted secrets, so moving from one to the other loses nothing. A BApp Store build loaded over a saved CLI backend runs on Burp AI and shows a notice; the saved choice is kept for the full build.

Building from source

Both builds need JDK 21:

The old single-JAR Gradle task is retired: running it fails with a message naming the two tasks above. See Installation for loading the JAR into Burp.

Last updated