h1-brain MCP
MCP server that gathers your HackerOne history and 3,600 disclosed reports into a database the agent uses to plan program work
- 356 ★ on the repository
- Last commit
Broad access
Why this level
- The HackerOne API token opens your reports, program scopes and conversations with security teams
- The briefing from hack puts the agent in offensive mode, and targets may be tested only within an authorized program
- The personal h1_data.db holds details of vulnerabilities you found, including ones not yet fixed
Install
- Works in
Manual install
claude mcp add h1-brain -e H1_USERNAME=your_hackerone_username -e H1_API_TOKEN=your_api_token -- /path/to/h1-brain/venv/bin/python /path/to/h1-brain/server.pyFirst clone the repository and install dependencies: git clone, python -m venv venv, pip install -r requirements.txt.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The HackerOne API token opens your reports, program scopes and conversations with security teams
- The briefing from hack puts the agent in offensive mode, and targets may be tested only within an authorized program
- The personal h1_data.db holds details of vulnerabilities you found, including ones not yet fixed
This is third-party code. Review the repository files before installing.
What it does
h1-brain is an MCP server for HackerOne bug bounty hunters. Through the API it pulls your rewarded reports, programs and their scopes into a local SQLite database. The agent searches it and reviews past findings by title, program, weakness type or severity. The repository ships a ready disclosed_reports.db with more than 3,600 publicly disclosed reports that paid a bounty, each with the vulnerability write-up, program and amount. The main tool, hack(handle), builds a program briefing in one call. It takes the fresh scope from the API, your past reports on that program and the weakness types that paid. It then finds bounty-eligible assets where you have no findings, checks public disclosures for the program and suggests lines of testing that paid on other programs. The briefing ends with agent instructions loaded from hack_instructions.md. The other tools fall into groups. For your database there are search_reports, get_report, get_report_thread and search_scopes, and for the public one search_disclosed_reports and get_disclosed_report. Sync is done by fetch_rewarded_reports, fetch_programs and fetch_program_reports, the last of which pulls all your reports on a program with their comment threads.
Who it is for. For security researchers in HackerOne programs who want the agent to know their past findings and public write-ups.
Good fit when
- You are starting on a program and want a briefing: scope, your past findings and assets you have not covered
- You want to see which weakness types paid on a particular program and compare them with your reports
- You want to search detailed public reports that paid a bounty, for example by weakness type or program name
- You want to export the whole history on a program, with comment threads and statuses, into markdown files
- Before writing a new report you want examples of similar disclosed reports
Not a fit when
- You have no HackerOne account and API token: without them the server gets neither your reports nor programs
- The target is outside the program scope: testing without the owner's permission may be unlawful and the server does not enforce this
- You need a scanner that runs checks against a target on its own: h1-brain prepares a briefing and sends nothing to the target
- You need data from other platforms such as Bugcrowd or Intigriti: the server works with HackerOne only
Example request
Call hack for the security program and show which bounty-eligible assets I have not covered yetMCP
- Transport
- stdio
- Authentication
- API key
Environment variables
H1_USERNAMErequired- HackerOne username
H1_API_TOKENrequired, secret- HackerOne API token from the settings page
Limitations
- Requires Python 3.10 or newer, a clone of the repository and a virtual environment. There is no PyPI or npm package, and the path to server.py is set by hand.
- A HackerOne API token and username are required. The token is issued in HackerOne settings and opens access to your reports.
- The public database holds only disclosed reports that paid and have a write-up: redacted and empty ones are excluded.
- The instructions in hack_instructions.md put the agent in offensive mode, so use the server only on programs where you have permission.
How to disable. Run claude mcp remove h1-brain in Claude Code or delete the h1-brain block from claude_desktop_config.json. Revoke the HackerOne token in the HackerOne settings. The local h1_data.db sits next to server.py; delete it if you no longer want personal reports stored.
FAQ
How do I connect h1-brain to Claude Code?
Clone the repository, create a venv and install the dependencies, then run claude mcp add h1-brain -e H1_USERNAME=login -e H1_API_TOKEN=token -- /path/to/venv/bin/python /path/to/server.py. For Claude Desktop the same block goes into claude_desktop_config.json.
What do I do after connecting?
Run fetch_rewarded_reports once: it is the most important step and loads your rewarded reports with write-ups. fetch_programs then pulls the programs you can access.
Where is my data stored?
Personal reports, programs, scopes and comment threads sit in the local h1_data.db file, which is created automatically and excluded from git. The public disclosed_reports.db ships with the repository. The server contacts HackerOne only with your token.
Can I export reports to markdown?
Yes. fetch_program_reports(handle, out_dir) pulls all your reports on a program in every state with their threads and saves each as a markdown file, plus raw JSON copies and an index.md file.
Do I need a HackerOne token to search the public reports?
No. The disclosed_reports.db database already ships in the repository, and you can search it right after connecting. The token is needed for your own reports, programs and scopes.
Discussion
No questions or reviews yet.
Related
MCP serversMCP server and CLI for reverse engineering: an agent inspects binaries, JavaScript apps, APKs and websites on your machine and returns the evidence
Nine skills that let an agent run Strix pentests, fix validated findings and add security scanning to CI
A skills router for authorized reverse engineering, pentest, CTF and threat analysis: it picks a methodology for the task and checks available tools
A code security audit skill by Cloudflare: the agent runs recon, coverage-led hunting and independent verification of findings, then produces a structured repor
A cybersecurity skill library: forensics, incident response, pentesting, cloud and malware analysis mapped to MITRE ATT&CK and NIST
NVIDIA's open stack for running OpenClaw, Hermes and LangChain Deep Agents in OpenShell sandboxes with network policy and managed inference
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| h1-brain (this entry) | no | API key | 356 | |
| Pentest Agent Suite | no | key is optional | 1,030 | a set of subagents and MCP servers for the testing itself, while h1-brain keeps history and prepares the briefing |
| MCP Kali Server | no | not required | 859 | runs nmap, sqlmap and other Kali tools on the agent's command |
README in short
The project describes how it exchanges data with HackerOne and how its two databases, personal and public, are built. For details the author points to a series of three articles on setup and work on a real program.