Not a search engine — it's precision routing. Each source card contains complete crawl strategies, anti-scraping configuration, and quality scoring.
Atlas generates structured source cards for every information source, including URLs, strategies, anti-scraping configuration, and quality scoring.
Instead of returning 10 links, it tells agents which site to use, what strategy to follow, and how to scrape. Every source card comes with a complete execution configuration.
High-performance search engine with BM25 ranking and quality scoring. Search latency is under 1 millisecond.
Agents automatically rate sources after use. The system learns which sources perform well — popular ones rise, poor ones are demoted. Self-optimizes hourly.
Every source card marks its anti-scraping level (green/yellow/red/black), automatically selects the optimal strategy, rotates user agents, and controls rate limits.
Covers Chinese, global, and academic information sources across multiple types. Filter precisely by region, language, and topic.
Call with one line of Python SDK code, access via REST API, or integrate through MCP Tools. Get your agent up and running in 5 minutes.
Add to your MCP config — point to the /mcp endpoint. 7 tools ready: search, list, get_source, contribute, stats.
Access the /mcp HTTP endpoint with any JSON-RPC client. Full tool set available without agent framework dependencies.
from runtime.sdk import search — one line to search. Built-in caching, retries, and quality scoring.
Add Atlas as an MCP server in your client config. 7 tools available for searching and managing information sources.