PentesterFlow is a new open-source, human-in-the-loop agentic AI command-line tool built specifically for penetration testers and bug bounty hunters, designed to automate recon-to-reporting workflows without sacrificing analyst oversight.
Most agentic AI security tools suffer from hallucinated findings, weak context retention, and poor tool integration, but PentesterFlow tackles these problems head-on with built-in pentest skills, evidence-based finding confirmation, and continuous local learning.
The tool connects to local or hosted large language models, plans actions against a scoped target, executes real offensive-security tools, and requires explicit analyst approval before running sensitive commands.
This “human-in-the-loop” design ensures the AI never runs wild on production systems without a security engineer’s sign-off, addressing a major trust gap in autonomous pentesting tools.
PentesterFlow
PentesterFlow is engineered to assist across the entire pentest lifecycle: scoping, reconnaissance, enumeration, validation, coverage tracking, reporting, and learning.
A live demo shows the tool loading a “webvuln” skill, sending HTTP requests to test an orders API, and automatically confirming a high-severity IDOR (Insecure Direct Object Reference) vulnerability, writing evidence-backed findings straight to a Markdown file.
| Feature | Description |
|---|---|
| Model backends | Ollama, LM Studio, Kimi, Groq, Gemini, DeepSeek, OpenRouter, and OpenAI-compatible APIs |
| Built-in skills | Recon, webvuln, SSRF, SSTI, JWT, GraphQL, race conditions, subdomain takeover, Supabase, deserialization |
| Tooling | Shell/Bash, HTTP, Burp Suite bridge, browser capture, MCP support, file read/write, grep/glob |
| Reporting | Confirmed findings saved with PoC, impact, remediation, and copy-pasteable curl commands |
| Memory | Session resume, context snapshots, and a continuous local intelligence system |
A notable feature is PentesterFlow’s local Continuous Learning System, which silently stores user preferences, successful workflows, coverage gaps, and lessons from failed assumptions in project-specific and personal intelligence files, improving future sessions without ever retraining the underlying model.
Secrets are automatically redacted before anything is written to disk, and duplicate memories are deduped before reaching the model, keeping the workflow simple even as it grows smarter over time.
PentesterFlow available from GitHub enforces permission-gated tool execution, blocks catastrophic shell command patterns, and redacts credentials during compaction and snapshotting, though it also offers a “YOLO mode” for auto-approving actions in isolated lab environments.
It integrates directly with Burp Suite via a companion bridge tool, letting testers send captured traffic into the CLI and import confirmed findings back as Burp issues for a smoother workflow between manual and AI-assisted testing.
Installation is straightforward via a shell script for macOS/Linux or a PowerShell command for Windows, pulling the latest standalone binary and verifying its SHA-256 checksum.
Users can pin specific versions, choose local LLMs like Ollama’s qwen2.5-coder model, or connect to hosted providers such as Gemini and Groq, then set a target URL with a simple /target command before instructing the agent in plain English, such as “test the orders API for broken access control.”
Analysts working with sensitive targets should note that PentesterFlow is explicitly intended for authorized security work only, given its ability to run shell commands and make live HTTP requests once approvals are granted.
As agentic AI tools proliferate across the offensive-security space alongside projects like PentAGI and PentestGPT, PentesterFlow’s emphasis on transparent, reproducible evidence and analyst-approved actions positions it as a notable entrant for teams wary of fully autonomous pentesting agents.
Strengthen Your SOC by Accelerating Threat Detection & Rapid Investigations. -> Integrate ANY.RUN With Your SOC Now.
