bob-ddg-mcp

A DuckDuckGo web search + document fetch MCP server for IBM Bob.

Gives Bob two live-internet tools:

• search — DuckDuckGo web search (no API key, no cost)
• fetch — Read any URL or PDF as clean text

Requirements

• Python 3.10 or later
• IBM Bob IDE

Install

git clone https://github.com/scprotz/bob-ddg-mcp.git
cd bob-ddg-mcp
python install.py

install.py will:

1. Create a dedicated .venv inside the project folder
2. Install dependencies (mcp, pypdf)
3. Verify the server starts correctly
4. Print the exact JSON snippet to add to Bob's MCP config

Configure Bob

After running install.py, add the printed snippet to your Bob MCP config.

Global (applies to all workspaces):

~/.bob/settings/mcp.json          # Windows: %USERPROFILE%\.bob\settings\mcp.json

Project-level (this workspace only):

.bob/mcp.json

Merge into the mcpServers object — the exact paths are printed by install.py and saved to mcp-config-snippet.json.

Then in Bob: Settings ⚙️ → MCP tab → refresh.

Usage

Just ask Bob naturally:

Search for IBM Storage Scale 5.2 release notes
Fetch https://example.com/doc.pdf and summarize it
What does this page say? https://www.ibm.com/support/pages/...

Bob will call search or fetch automatically as needed.

Tool reference

Tool	Parameters	Description
search	query (str), max_results (int, default 5)	DuckDuckGo web search
fetch	url (str), max_chars (int, default 20000)	Fetch HTML page or PDF as text

Updating

cd bob-ddg-mcp
git pull
python install.py   # re-runs if dependencies changed

Or just git pull — if only ddg_search.py changed, no reinstall is needed. Bob hot-reloads the script.

Changelog

See CHANGELOG.md.

Notes

• DuckDuckGo does not have an official public API. This server uses their unofficial search endpoint, which works reliably for personal/productivity use.
• No data is sent to any third party beyond DuckDuckGo and the URLs you fetch.
• All processing is local — the venv runs on your machine as a Bob child process.