[georgeatparallel]

Summary

Give the CMO and specialist agents free, browser-free web research with Parallel Search MCP. Set OPENCMO_SEARCH_PROVIDER=parallel for Fast-mode search without an account or search API key. Source URLs and excerpts go back to the agent. Unset/auto keeps existing routing; none disables this tool. Explicit Parallel failures do not switch providers. SERP tracking and community scans keep their existing integrations.

On Artificial Analysis's Search API benchmark, Parallel Fast records 0.802 DeepSearchQA answer-item F1 versus Tavily Basic's 0.744. Benchmark source.

I work at Parallel.

Verification

• Commands run: pytest tests/test_parallel_search.py tests/test_tavily_mcp.py tests/test_smoke.py tests/test_crawl.py tests/test_browser_pool.py (39 passed on both minimum and current Agents SDK versions); ruff check src/ tests/.
• Manual checks: anonymous search through the CMO's registered tool returned source URLs and excerpts.
• Screenshots or notes: backend change; setup documented in the configuration guide.

Checklist

• The change is focused and scoped to one topic
• Relevant tests and builds were run
• Docs were updated if behavior or configuration changed
• My commits use an email linked to my GitHub account or my GitHub noreply email
• If I used AI or automation, human attribution is preserved via the commit author or Co-authored-by