AI-Driven Offensive OperationsBeyond The Scan
From repository onboarding and attack-surface discovery to live engagements, remediation, sidecars, and MCP-powered AI orchestration
Target Repository Access
Install the Reaper GitHub App for fine-grained, tokenless repository access. Automatic webhook-driven scans on every push and pull request event. Full OAuth installation flow, so there is no personal access token to rotate.
- GitHub App - One-click OAuth installation with fine-grained permissions
- Webhook Events - Automatic scans on push, PR, and installation events
- Tokenless Access - Short-lived JWT tokens, no PATs to rotate
- Enterprise Ready - GitHub.com and GitHub Enterprise Server support
- PAT Fallback - Personal Access Token auth for other VCS platforms
Dependency Inventory From Target Repos
Reaper parses the dependency manifests in a cloned target repository and pins exact versions from lock files where they exist, so a CVE match names a version rather than a range.
go.mod/go.sum- Go modulespackage.json/yarn.lock/pnpm-lock.yaml- Node.js (npm, yarn, pnpm)requirements.txt/Pipfile.lock- PythonCargo.toml/Cargo.lock- Rustpom.xml/build.gradle- JavaDockerfile/docker-compose.yml- Container images & packagesMakefile/.mise.toml- Tool versions
Multi-Source Exploit Database
Aggregated exploit intelligence from four offensive security databases. Know which CVEs have working exploits, which are being actively exploited in the wild, and which have Metasploit modules ready to deploy.
- Exploit-DB - Curated public exploits with PoC code from Offensive Security
- GitHub PoCs - CVE proof-of-concept exploits collected from GitHub repositories
- InTheWild - Real-time tracking of vulnerabilities actively exploited in the wild
- Metasploit Modules - CVE-to-module mappings from the Metasploit Framework
- Full-text search across all exploit sources with CVE correlation
- Continuous synchronization with configurable source selection
Operation Notifications
Get notified the moment a new vulnerability affects your tracked versions. Configure alert rules by severity, ecosystem, or project.
- Email notifications
- Slack webhooks
- Configurable severity thresholds
- Project-specific rules
express@4.17.1Project Control Rooms And Reporting
Every project gets a deep workspace for alerts, vulnerabilities, logs, AI review, Semgrep coverage, access control, and reporting. Export project and organization reports as PDF or Markdown.
- Per-Project Views - Alerts, vulnerabilities, logs, history, AI audit, code, access, and settings from a single screen
- Executive Reports - Export PDF or Markdown reports with stakeholder-ready summaries
- Guardian Visibility - Review deterministic sidecar output next to offensive findings
- Email Delivery - Send generated reports directly to stakeholders
Vulnerability Scanning
Offensive scanning combines native tooling with optional sidecars. Tune tags, severities, scripts, and execution scope while mixing live recon with code and dependency signal.
- Nuclei - Configurable template tags (cve, rce, sqli, xss, lfi, ssrf), severity filters, custom template directories
- Nmap - NSE script categories (vuln, exploit, auth, brute, discovery), port selection, timing templates, full script output parsing
- Core Tooling - ffuf, subfinder, httpx, katana, semgrep, detect-secrets, grype, trivy
- Guardian Sidecar - Deterministic code, dependency, CVE, and license analysis merged into the offensive pipeline
- Kali Sidecar - Optional isolated execution for allowlisted tools such as hydra, nikto, wpscan, gobuster, and hashcat
- CWE and CVE correlation with exploit database cross-referencing
- Parallel scanning with smart deduplication and audit trails
Exploit Review, Reporting, And Remediation
Use AI to review exploitability, generate proof-of-concepts, write reports, and open remediation work without losing operator control.
- Finding Review - AI analyzes pentest findings to identify false positives vs true vulnerabilities
- CVE Impact Assessment - Evaluate real-world exploitability based on attack vectors and context
- PoC Generation - Generate proof-of-concepts for vulnerabilities and scan findings
- Security Report Generation - Draft incident and scan reports with remediation steps attached
- Fix PR Workflows - Create and track AI-generated remediation pull requests
- Impact Summaries - Executive-level summaries explaining security posture in plain language
AI Orchestration
A full Model Context Protocol server with query, analysis, and action workflows. Let Claude Code, Cursor, Windsurf, or any MCP-compatible AI assistant directly query vulnerabilities, search exploit databases, trigger scans, create fix PRs, and manage your security posture conversationally.
- 18 Query Tools - Explore org overview, project details, dependencies, vulnerabilities, pentest findings, alerts, CVE info, exploit database search, exploits by CVE
- 6 Analysis Tools - PoC generation, license compliance, security report generation, report export
- 11 Action Tools - Create fix PRs, trigger scans, manage alerts, create incidents, import projects
- Guided Prompts - Security triage, project review, and license compliance audit workflows
- Auth Options - Per-org enablement with RBAC, API keys, and OAuth 2.1 for remote MCP clients
Deep Recon Agent
A dedicated AI agent walks the source with code-exploration tools using a 3-phase methodology. Finds logic flaws, race conditions, and insecure design patterns that automated scanning tools cannot detect.
- Phase 1: Reconnaissance - Maps project structure, identifies entry points, security-critical areas
- Phase 2: Code Analysis - Examines auth, access control, data handling, cryptography, database access
- Phase 3: Logic Review - Business logic flaws, race conditions, TOCTOU bugs, insecure design patterns
- Configurable Focus - Target auth, injection, crypto, data exposure, logic, or all areas
- Fix PR Generation - AI generates pull requests to fix discovered audit findings
AI Exploit Agent
Every Scan finding and CVE alert is reviewed by AI for real-world exploitability. The system clones your repo, analyzes actual code paths, and determines whether vulnerable functions are truly reachable -- so the queue holds findings an attacker could actually reach.
- Relevance Scoring - 0.0 to 1.0 score based on code reachability and attack vector analysis
- False Positive Detection - Identifies test files, dead code, sanitized inputs, framework protections
- Dependency Usage Analysis - Checks if vulnerable functions are actually called in your code
- PoC Generation - Creates proof-of-concept exploits to validate true threats
- Auto Review - Automatically reviews all critical/high findings during scans
AI-Generated Semgrep Rules
A multi-agent AI pipeline generates custom Semgrep rules tailored to your exact codebase. Organization-level rules from natural language descriptions, plus project-specific rules that target your frameworks, APIs, and security patterns.
- 4-Agent Pipeline - Analysis, Implementation, Verification, and Fix agents collaborate to produce valid rules
- Project-Specific Rules - AI analyzes your codebase to generate 3-8 targeted rules per project
- Per-Project Config - Granular control: enable/disable rule groups, cherry-pick individual rules, override group defaults
- Batch Generation - Describe complex security patterns in natural language, AI breaks them into individual rules
- Auto-Validation - Every generated rule is validated with
semgrep --validateand auto-fixed if needed
Domains, Hosts, And Continuous Discovery
Monitor your external attack surface with passive DNS-based discovery. Automatically find subdomains, track live domains, and maintain visibility into all your internet-facing assets.
- Passive DNS Discovery - Find subdomains without active scanning using historical DNS data
- Live Domain Probing - Automatically check which discovered domains are active and responding
- IP Resolution - Resolve and track IP addresses for live domains
- Host Tracking - Maintain an inventory of all hosts associated with your projects
- Continuous Monitoring - Scheduled discovery and probing keeps your asset inventory current
Scheduled Operations And Incident Response
Run recurring engagements, inspect execution trees, and convert validated findings into incidents with AI-assisted reporting.
- Engagement Scheduling - Create recurring offensive runs and trigger them on demand
- Execution Trees - Inspect per-task progression, live logs, ETA, and operator checkpoints
- Incident Lifecycle - Track incidents through open, investigating, mitigating, resolved, and closed states
- Alert Linking - Connect vulnerability alerts to incidents for full context
- Timeline Tracking - Maintain a detailed timeline of all incident actions and updates
- AI Bootstrap - Create incidents from URLs, free text, or PDFs
- AI-Enhanced Reports - Write up incidents with the remediation steps for each finding
Identity, Governance, And Access Control
Manage collaboration at scale with multi-tenant organizations, RBAC, team-based access, and modern identity controls for operators and AI clients.
- Multi-Tenancy - Isolated organizations with separate projects, users, and settings
- Team Management - Organize users into teams with leads and members
- Role-Based Access - Admin, member, and viewer roles with granular permissions
- Project Access Control - Control which teams can access specific projects
- Identity Options - Passkeys, API keys, OAuth apps, SSO, and SCIM provisioning
Dedicated, Managed Deployments
The standard Reaper offering is a dedicated web-service deployment for each client, managed by Amsterdam Technologies. Deployment on customer infrastructure or on-premises is available when required.
- Dedicated Environment - An isolated web-service deployment for each client
- Managed Operations - Deployment and platform operations handled by Amsterdam Technologies
- Full Platform - UI, auth, org settings, scheduling, and reporting
- Customer Infrastructure - Optional deployment in your cloud or on-premises environment
- Guardian - Optional deterministic defensive scanning sidecar
- Kali - Optional isolated offensive sidecar with allowlisted remote execution
Feature Matrix Against OSS Pentest Agents
Repository names link directly to their upstream GitHub projects.
| Feature | Reaper | PentAGI | Strix | Shannon |
|---|---|---|---|---|
| Full web platform / multi-tenancy | YesOrgs, teams, RBAC, projects, auth | PartialWeb UI and APIs; tenancy differs | NoOSS CLI; platform separate | NoLite CLI/workspaces; Pro commercial |
| Autonomous pentest pipeline | YesScan, exploit, review, report phases | YesFlow to task hierarchy | YesDynamic root/sub-agent graph | YesTemporal 5-phase workflow |
| Defensive SAST / SCA / CVE integration | YesGuardian sidecar plus Semgrep/SCA pipeline | PartialSecurity tooling, not Guardian equivalent | PartialDynamic-first app testing | PartialLite source analysis; SCA in Pro |
| Live offensive tool execution | Yes39 scoped engagement tools | Yes20+ sandboxed security tools | YesTerminal, Python, proxy, browser tools | YesBrowser and bounded CLI tools |
| Browser / proxy sandbox | NoNo Caido/Playwright agent sandbox yet | PartialBrowser scraper; no Caido-style proxy | YesCaido, Playwright, terminal sandbox | PartialPlaywright sessions, no proxy layer |
| MCP server / AI-assistant integration | Yes33 MCP tools plus engagement catalogue | NoREST/GraphQL, no MCP server found | NoOwn CLI/runtime tools | NoClaude SDK, not MCP server |
| Multi-provider LLM | YesGemini, Anthropic, OpenAI, OpenRouter, Claude CLI | Yes10+ providers | YesLiteLLM providers | PartialClaude via Anthropic, Bedrock, Vertex, custom endpoint |
| Offline / local LLM | PartialOpenAI-compatible base URL; no first-class Ollama/offline mode | YesOllama/vLLM support | YesOllama/LMStudio via LiteLLM | NoNon-Claude/local models not officially supported |
| Semantic memory / knowledge | YesMongoDB vector search when embeddings are available | Yespgvector plus Graphiti | PartialWiki notes, tag lookup, compression | NoWorkspaces/logs, not semantic memory |
| Cost budget enforcement | YesPer-run and daily USD caps cancel runs | NoLimits and observability; no spend cap found | PartialCost tracking; no hard cap found | YesSpending cap detection |
| Operator checkpoints | YesDecision cards with SSE and REST fallback | PartialMentor/refiner supervision | PartialInteractive CLI/TUI | PartialCLI-controlled workflow |
| Fix PR / remediation workflows | YesAI remediation and VCS fix PR flow | PartialDeveloper agent can generate code | YesOne-click fix PRs | NoLite reports; Pro separate |
| Reporting / export | YesMarkdown, PDF, OSCP skills, emails | YesFlow reports and logs | YesReports and compliance output | YesComprehensive markdown report |
| Enterprise identity / governance | YesJWT, API keys, OAuth, SSO, SCIM | PartialAuth and API tokens; governance differs | NoOSS CLI; enterprise separate | NoLite CLI; Pro separate |
Download the Reaper Whitepaper
Offensive security platform covering attack-surface discovery, live engagements, AI-powered auditing, remediation, and MCP integration.