yeswehack-mcp
# yeswehack-mcp
A [Model Context Protocol (MCP)](https://modelcontextprotocol.io) server for the [YesWeHack](https://yeswehack.com) bug bounty platform. Lets Claude query your private and public programs, reports, and the hacktivity feed directly from a conversation.
## Features
- **Flexible authentication** — browser login, email/password API login, copied bearer tokens, and official Personal Access Tokens.
- **Private programs** — returns invite-only programs you have been accepted into, not just public ones.
- **Full program details** — scope, reward ranges, status.
- **Report access** — list and read reports with severity, CVSS, description, and bounty.
- **Report comments** — list report discussion/messages when your account has access.
- **Email aliases** — list your YesWeHack email aliases.
- **Program credentials** — list credential pools/assigned credentials and request credentials for programs that expose pools.
- **Hacktivity feed** — browse publicly disclosed reports.
- **Token caching** — the JWT is stored locally and reused until it expires.
## Requirements
- Python 3.10+
- [uv](https://docs.astral.sh/uv/) (installed automatically by the setup script if missing)
- A YesWeHack account
- **WSL2 users:** WSLg must be enabled so Chromium can open a window (`echo $DISPLAY` should return a value)
## Installation
```bash
git clone https://github.com/youruser/yeswehack-mcp
cd yeswehack-mcp
# Install uv if you don't have it
curl -LsSf https://astral.sh/uv/install.sh | sh
# Install dependencies
uv sync
# Download the Chromium browser used for login
uv run playwright install chromium
```
## Registering with Claude Code
```bash
claude mcp add yeswehack -- uv --directory /path/to/yeswehack-mcp run server.py
```
Replace `/path/to/yeswehack-mcp` with the actual path where you cloned the repo.
Verify it connected:
```bash
claude mcp list
# yeswehack: ... ✓ Connected
```
## Registering with Claude Desktop
Add the following to your Claude Desktop config file:
- **macOS:** `~/Library/Application Support/Claude/claude_desktop_config.json`
- **Windows:** `%APPDATA%\Claude\claude_desktop_config.json`
```json
{
"mcpServers": {
"yeswehack": {
"command": "uv",
"args": [
"--directory",
"/path/to/yeswehack-mcp",
"run",
"server.py"
]
}
}
}
```
If `uv` is not on PATH when Claude Desktop launches, use the absolute path (e.g. `/home/youruser/.local/bin/uv`).
## Usage
Once registered, start every session by authenticating:
> **You:** Call the authenticate tool
With no arguments, a Chromium window opens. Log in to YesWeHack as normal (email + password + 2FA if enabled). The window closes automatically once your session is detected. The token is saved to `~/.config/yeswehack-mcp/token.json` and reused for all subsequent calls until it expires.
You can also authenticate without a browser:
```text
authenticate(email="you@example.com", password="...", totp="123456")
authenticate(access_token="eyJ...") # browser/API bearer token
authenticate(access_token="ywh_pat_...") # Personal Access Token
```
Environment variables are supported too:
```bash
export YWH_TOKEN="eyJ..." # browser/API bearer token
export YWH_PAT="ywh_pat_..." # Personal Access Token
```
### Available tools
| Tool | Description |
|------|-------------|
| `authenticate` | Browser login, API login, or store a copied bearer token/PAT |
| `get_current_user` | Show your YesWeHack profile (username, rank, reputation) |
| `list_programs` | List all programs you have access to, including private invite-only ones |
| `get_program` | Full details for a program: scope, reward ranges, status |
| `list_reports` | List reports for a program, with optional status filter |
| `get_report` | Full details of a specific report (title, severity, CVSS, description, bounty) |
| `list_report_comments` | List comments/messages for a report when your token has access |
| `list_email_aliases` | List your YesWeHack email aliases |
| `get_program_credentials` | List credential pools and assigned credentials for a program |
| `request_program_credentials` | Request credentials from a program credential pool |
| `yeswehack_api_get` | Read-only escape hatch for authenticated API endpoints not wrapped yet |
| `get_hacktivity` | Browse the public hacktivity (disclosed reports) feed |
### Example prompts
```
List all my private programs on YesWeHack.
Show me the scope for the program with slug "acme-corp".
List my YesWeHack email aliases.
Get credentials for program "acme-corp".
List all accepted reports for program "acme-corp".
Get the full details of report 12345.
Show me the latest hacktivity, page 2.
```
## Token storage
The token is saved to `~/.config/yeswehack-mcp/token.json`. It contains only the token and its expiry timestamp — no account password is ever stored. Browser/API session JWTs use their embedded expiry. Opaque Personal Access Tokens are cached locally until the YesWeHack API rejects them. The browser profile (cookies, localStorage) is kept at `~/.config/yeswehack-mcp/browser-profile` so you do not have to fill in your email every time you re-authenticate.
To log out, delete the token file:
```bash
rm ~/.config/yeswehack-mcp/token.json
```
## Project structure
```
yeswehack-mcp/
├── server.py # FastMCP entry point — all tool definitions
├── auth.py # Playwright browser login + token storage
├── client.py # httpx async API wrapper with pagination
└── pyproject.toml # Dependencies and build config
```
## WSL2 display setup
Playwright needs a display to open the browser window. On WSL2 with WSLg this works out of the box. If you see an error about `DISPLAY` not being set:
```bash
# Check if WSLg is running
echo $DISPLAY # should print something like :0
ls /mnt/wslg # should exist
# If not, ensure you are on a recent WSL2 version with WSLg support
wsl --update # run from Windows PowerShell
```
## License
MIT
TDQS
Scored across 12 tools
Most tools pair a distinct resource with a clear verb (list_programs vs get_program, list_reports vs get_report, get_program_credentials vs request_program_credentials). The main ambiguity is yeswehack_api_get, a generic read-only escape hatch that conceptually overlaps with all dedicated getters, though its 'unwrapped endpoints only' framing reduces confusion.
The dominant verb_noun pattern is consistent (list_programs, get_report, request_program_credentials), but yeswehack_api_get reverses the order with a service prefix, and authenticate is a lone verb without a noun. These deviations are noticeable but not chaotic.
12 tools is well-scoped for a bug bounty platform client: authentication, programs, reports, comments, credentials, user profile, hacktivity, and one escape hatch. Each tool earns its place and the count is neither bloated nor thin.
The read surface is solid (programs, reports, comments, credentials, hacktivity), but state-changing operations central to the bug bounty workflow are missing: there is no way to submit a vulnerability report or post a comment on a report. The read-only escape hatch cannot fill these gaps, so researchers cannot complete a full workflow.