Nimble Search

nimble search — reference

Real-time web search with 8 focus modes. Returns results with titles, URLs, and optionally full content and AI answers.

Table of Contents

• Parameters
• Search depth modes
• Focus modes
• CLI
• Python SDK
• Response structure

Parameters

CLI uses hyphens (--search-depth, --include-answer). SDK uses underscores (search_depth, include_answer).

Search depth modes

Default for AI agent use: prefer fast — richest content-to-latency ratio.

Focus modes

CLI

# Fast depth — rich content, low latency (best for agents)
nimble search --query "React server components" --search-depth fast

# Lite — metadata only, fastest
nimble search --query "OpenAI announcements" --focus news --search-depth lite

# Deep — full real-time page scrape
nimble search --query "EU AI Act" --focus news --search-depth deep \
  --start-date 2025-01-01 --end-date 2025-12-31

# With AI answer + domain filter
nimble search --query "Python asyncio best practices" \
  --focus coding --search-depth fast --include-answer \
  --include-domain '["docs.python.org", "realpython.com"]'

# Extract just URLs
nimble --transform "results.#.url" search --query "React tutorials" --search-depth lite

Python SDK

from nimble_python import Nimble
nimble = Nimble(api_key=os.environ["NIMBLE_API_KEY"])

# Fast depth — best default for AI agent use
resp = nimble.search(query="React server components", search_depth="fast")

# Lite — scan many results quickly
resp = nimble.search(
    query="OpenAI announcements",
    focus="news",
    search_depth="lite",
    time_range="week",
)

# Deep — full content for research
resp = nimble.search(
    query="EU AI Act enforcement",
    focus="news",
    search_depth="deep",
    include_answer=True,
)

# Custom focus — explicit agent array
resp = nimble.search(
    query="best wireless headphones",
    focus=["amazon_serp", "walmart_serp"],
    search_depth="fast",
    max_results=10,
)

results = resp.results       # list of result objects
answer = resp.answer         # AI summary (if include_answer=True)

Response structure