burp-mcp（burp MCP 效率）

Burp MCP

Use this skill to talk to a local Burp Suite MCP Server exposed by the PortSwigger extension.

When to use

• Burp Suite is already running locally
• The MCP extension is loaded and enabled
• You want to inspect available Burp tools
• You want to call a specific Burp MCP tool from the terminal or from OpenClaw via exec

Endpoint

Default endpoint used by this skill:

http://127.0.0.1:9876/

Important: on this machine, the live SSE endpoint is /, not /sse.

Commands

List Burp tools:

python ./skills/burp-mcp/scripts/burp_mcp.py list-tools

Call a Burp tool:

python ./skills/burp-mcp/scripts/burp_mcp.py call <tool_name> '<json_args>'

Examples:

python ./skills/burp-mcp/scripts/burp_mcp.py list-tools
python ./skills/burp-mcp/scripts/burp_mcp.py call get_proxy_http_history '{"offset":0,"count":5}'
python ./skills/burp-mcp/scripts/burp_mcp.py call get_proxy_http_history_regex '{"offset":0,"count":10,"regex":"login|token|auth"}'
python ./skills/burp-mcp/scripts/burp_mcp.py call output_project_options '{}'

How agents should use it

1) Always start with list-tools

Burp MCP tool names and schemas can change with extension version. Before calling tools in a new setup, list tools first:

python ./skills/burp-mcp/scripts/burp_mcp.py list-tools

This returns JSON entries with:

• name
• description
• inputSchema

2) Match args to inputSchema

When calling a tool, build the JSON object to match the tool's required schema exactly.

Example:

• get_proxy_http_history requires:

- offset (integer) - count (integer)

So call it like:

python ./skills/burp-mcp/scripts/burp_mcp.py call get_proxy_http_history '{"offset":0,"count":20}'

3) Prefer safe read-only tools first

Safe first calls:

• get_proxy_http_history
• get_proxy_http_history_regex
• get_proxy_websocket_history
• get_scanner_issues
• output_project_options
• output_user_options

Be careful with mutation tools such as:

• set_project_options
• set_user_options
• set_proxy_intercept_state
• set_task_execution_engine_state
• set_active_editor_contents

Output behavior

• list-tools prints JSON array of tools
• call prints JSON object:

- isError - content

Burp often returns tool results as text blocks. Some tools return newline-separated JSON strings inside content[].text, so callers may need a second parse step if they want structured output.

Troubleshooting

• If connection fails, verify Burp is open and the MCP extension is enabled
• If /sse fails, use / instead
• If tool calls return schema errors, re-run list-tools and check inputSchema
• If nothing returns, test manually with:

curl http://127.0.0.1:9876 -v

A healthy server should return text/event-stream

Notes

• Uses the official Python mcp SDK
• Requires asyncio
• Assumes the mcp package is already installed in the active Python environment