Breachwright

Breachwright is open-source penetration test management software for organizing assessment evidence, turning scanner output into reviewable findings, mapping attack paths, and producing client-ready reports.

Created by Advent Cybersecurity and released as open source for the security community.

Open-source release

Breachwright has one distribution with the complete product feature set. There are no paid editions, activation keys, seat limits, engagement limits, finding limits, feature gates, or subscription checks.

The former hosted Advent AI provider is not part of the open-source release because it depended on Advent-operated infrastructure and a private access service. Every AI-assisted product workflow remains available through user-controlled providers:

• Anthropic
• OpenAI
• Azure OpenAI
• AWS Bedrock
• Ollama, vLLM, llama.cpp, LM Studio, and other compatible local endpoints

Third-party AI services may charge for API usage. Local model support does not require a commercial API.

Quick start

1. Open the latest release and download the archive for Windows x64 or Linux x64.
2. Verify the archive against the release's SHA256SUMS.txt file.
3. Follow INSTALL.md to extract and install the complete package.
4. Launch Breachwright and create an engagement. No account or activation is required.
5. Import authorized scanner output or capture a note in the Evidence Notebook, review the resulting evidence, and promote validated work into findings.
6. Use the engagement Overview to track coverage, retests, readiness, and reporting.

See the workflow

Features

• A refreshable engagement Overview with readiness, retest priorities, recent local activity, and direct access to the next assessment task
• Conservative local auto-detection and scan ingestion for Nmap, Nessus, Burp Suite, Nuclei JSONL, SARIF, and raw evidence
• Provider-free cross-tool correlation with bounded file-count, per-file, and combined input limits
• A latest-snapshot asset and service inventory with new, persistent, regressed, and resolved observation context
• Engagement-wide local search across findings, checklist items, assets, evidence metadata, notebook notes, and exploitation chains
• An Evidence Notebook for raw analyst notes and attachments before they are ready to become findings
• Explicit no-AI promotion of scanner observations and reviewed notebook notes into findings with retained provenance
• Explicit per-scan AI selection plus a provider-free preflight for input size, provider choice, local redaction state, and readiness
• AI-assisted finding drafts with severity, CVSS, evidence, and remediation
• Provider, local-redaction, and potential-cost notices before every generative action, with fail-closed privacy checks and explicit confirmation when redaction is disabled
• Evidence-grounded AI review with source excerpts, confidence, create/update diffs, and accept, edit, reject, or bulk review controls
• Versioned scan snapshots with deterministic retest comparison
• Finding change history, due and overdue retest work, recently remediated review, and report readiness checks
• Built-in and user-created versioned engagement templates with automatic methodology checklists
• Reusable versioned finding templates that exclude target-specific hosts and evidence
• Current OWASP Top 10:2025 coverage for web engagements
• Dedicated OWASP API Security Top 10 (2023) checklist for API engagements
• Nuclei JSONL, SARIF 2.1, and spreadsheet-safe CSV interoperability with optional local redaction
• Versioned engagement export and import that preserves checklist progress, finding history, and normalized scan comparison history
• Exploitation chains and MITRE ATT&CK-aware attack narratives
• Bounded SharpHound and BloodHound ZIP import with Active Directory attack-path analysis and explicit dataset deletion confirmation
• Markdown and DOCX report generation
• Provider-free AI report preflight with context limits, redaction visibility, external-provider cost notice, and deterministic local fallback
• Verified local backup and offline restore with secret exclusion
• Built-in database and stored-file integrity diagnostics, backup freshness, and version visibility
• Privacy-bounded support snapshots for community troubleshooting without logs, credentials, local data paths, or assessment content
• Validated image, PDF, HTTP, request, response, HAR, text, Markdown, CSV, and JSON evidence attachments
• Tool Runner workflows for nmap, nikto, subfinder, feroxbuster, nuclei, and related tools, with completed output reusable in Scans or the Evidence Notebook
• PTES, OWASP, and NIST methodology checklists and gap analysis
• Cross-engagement intelligence, client risk profiles, and recurring-finding analysis
• Custom report templates and AI prompts
• Immediate access to one local owner workspace with no account or login setup
• Light and dark themes

OWASP Top 10 checklists are practical baselines, not claims of complete test coverage. Use a scope-appropriate verification standard and methodology for the full assessment.

Local application security

Breachwright can execute security tools and process sensitive assessment data. Use it only on systems and data you are authorized to test. Review Tool Runner commands before execution, protect the application data directory, and do not expose the API directly to untrusted networks.

Breachwright is designed as a single-owner local desktop tool. It opens directly into the workspace without accounts, passwords, roles, or sessions. The packaged application binds only to the local machine. Anyone who can use the operating-system account can access Breachwright data and run its tools, so rely on workstation login, disk encryption, and file permissions.

See SECURITY.md for vulnerability reporting and supported versions.

Install from source

Requirements

• Python 3.11 or newer
• Node.js 20 or newer
• GTK3 and WebKit2 on Linux when using the desktop window
• Edge WebView2 on Windows when using the desktop application
• Any external assessment tools you want to invoke through Tool Runner

Linux and macOS

git clone https://github.com/Advent-Cybersecurity/breachwright.git
cd breachwright
python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r backend/requirements.txt
cd frontend
npm ci
npm run build
cd ..
python run.py

Windows PowerShell

git clone https://github.com/Advent-Cybersecurity/breachwright.git
Set-Location breachwright
py -3.11 -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install -r backend\requirements.txt
Set-Location frontend
npm ci
npm run build
Set-Location ..
python run.py

On first launch, Breachwright opens directly into the local workspace.

Windows and Linux packages

Breachwright uses one source tree, one feature set, and one data format on Windows and Linux. Native desktop dependencies differ, so release candidates are produced as two archives:

• Windows x64 ZIP with Breachwright.exe and BreachwrightCLI.exe
• Linux x64 tar.gz with Breachwright and BreachwrightCLI

Both archives are built and tested by the same candidate workflow. A release download is not updated until both native candidates pass their platform checks. See INSTALL.md for extraction, installation, first-run, data-location, backup, and uninstall instructions.

The Windows executables are not currently Authenticode-signed. Windows may show an unknown-publisher or Microsoft Defender SmartScreen warning. Download only from the official Advent Cybersecurity release, verify the published SHA-256 checksum before running the package, and do not proceed if it differs.

Configure an AI provider

The application stores configuration in its platform-specific data directory:

• Windows: %APPDATA%\Breachwright\.env
• macOS: ~/Library/Application Support/Breachwright/.env
• Linux source or direct bundle: ${XDG_DATA_HOME:-~/.local/share}/breachwright/.env
• Linux installed package: ${XDG_DATA_HOME:-~/.local/share}/breachwright/data/.env

You can configure Anthropic, OpenAI, Azure OpenAI, AWS Bedrock, or a compatible local endpoint in the Settings page. Start with .env.example when configuring through an environment file instead.

Anthropic and OpenAI use a tested Breachwright recommendation by default, so their normal setup requires only the provider and API key. Settings shows the effective model, and an Advanced section allows an exact model override. Azure OpenAI still requires its endpoint and deployment name. AWS Bedrock requires a region plus a model or inference profile available to that AWS account. Local servers require their URL and an installed model, which Settings can detect.

Common credential patterns are redacted locally before AI context is sent by default. The setting is visible and can be changed by the operator. Redaction reduces accidental disclosure but is not a guarantee that every sensitive value will be detected, so review assessment data and provider terms before using an external model.

Before scan analysis starts, the Scans page shows the exact selected file count, combined size, configured provider, and redaction state without calling that provider. Select only the uploads needed for the task. Other stored scans remain available for evidence and snapshot work.

AI configuration is optional for manual findings, evidence management, checklists, reporting from existing content, export/import, and other non-AI workflows.

Engagement JSON exports are intended for sharing editable project records and normalized comparison history. They do not include raw scan files, finding or notebook attachment files, notebook notes, Active Directory datasets, pending AI proposals, generated reports, or Tool Runner output. Use a verified full backup when moving or preserving an entire local workspace.

AI output is treated as untrusted. Scan and Active Directory analysis create review proposals rather than accepted findings. Each supported proposal cites stored evidence, and nothing enters the Findings list until the local operator accepts it. See docs/AI_TRUST_AND_EVALUATION.md.

Back up and restore data

Create and download verified local backups from the Settings page. Backups include the SQLite database, finding and notebook attachments, uploads, reports, custom template assets, and Tool Runner output. API keys and environment configuration are excluded.

Restores are offline by design and preserve displaced data in a recovery folder. See docs/DATA_SAFETY.md for packaged and source commands.

Docker

The included Docker Compose configuration runs the API, PostgreSQL, and the web frontend:

cp .env.example .env
# Set POSTGRES_PASSWORD in .env before continuing.
docker compose up --build

The Docker deployment listens only on 127.0.0.1:80 by default. Application files persist in ./data; PostgreSQL data persists in the pgdata volume. The built-in portable backup currently supports SQLite installations, so Docker/PostgreSQL users must back up both PostgreSQL and ./data. Do not change the loopback binding unless you add a separate, deliberate access control and transport-security layer.

Architecture

• Backend: FastAPI, SQLAlchemy, Alembic, SQLite or PostgreSQL
• Frontend: React, Vite, Tailwind CSS
• Desktop: pywebview
• Reports: python-docx and Markdown
• Packaging: PyInstaller for Windows and Linux

See the documentation index for architecture, data safety, AI trust, dependency security, and release records.

Contributing

Issues and pull requests are welcome. Read CONTRIBUTING.md before submitting a change. Community support expectations are described in SUPPORT.md.

The project direction, cross-platform release standards, and cost and privacy guardrails are documented in ROADMAP.md.

License and attribution

Breachwright is licensed under the Apache License 2.0.

Copyright 2026 Advent Cybersecurity LLC.

The Apache License permits use, modification, and redistribution under its terms. It does not grant permission to use Advent Cybersecurity trademarks or imply endorsement. See NOTICE, TRADEMARKS.md, and THIRD_PARTY_NOTICES.md.