RUC-NLPIR

RUC-NLPIR / SearchClaw

Public

Harness for deep search agent

github.comRUC-NLPIRSearchClawblobmainreportSearchClaw.pdf
41
1
100% credibility
Found Apr 12, 2026 at 41 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

SearchClaw is a self-hosted web research agent that autonomously searches the web, academic papers, news, and other sources to generate well-cited answers to user questions.

How It Works

1
🔍 Find SearchClaw

You discover SearchClaw, a helpful tool that does deep research for you by searching the web and pulling together answers with trustworthy sources.

2
đź’» Set it up on your computer

Download and install it easily on your own computer so it runs privately just for you.

3
🧠 Connect smart helpers

Link it to thinking services and search helpers so it can understand questions and find information.

4
🚀 Launch your research buddy

Start the tool with a simple command, and it opens a friendly web page in your browser ready to help.

5
đź’­ Ask your question

Type any question you have, like 'What's the latest on climate change solutions?', and hit go.

6
🔬 Watch it research live

See it search the web, read articles, check papers and news, and build a complete picture step by step.

âś… Get your trusted answer

Receive a clear, detailed response with links to all the sources so you can verify and learn more confidently.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 41 to 41 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is SearchClaw?

SearchClaw is a self-hosted Python research agent that runs as a FastAPI web server with a browser UI. You ask a question, and it autonomously searches the web, academic papers, news, and WeChat; fetches pages (with optional browser support for JS sites); then synthesizes a well-cited answer. Drawing from harness engineering deep agents principles—like quality gates and task decomposition—it solves unreliable LLM prompting by enforcing thorough, verifiable research in a local setup.

Why is it gaining traction?

Unlike basic RAG tools or plain chatbots, SearchClaw's agentic harness rejects incomplete answers via citation checks, source diversity rules, and plan completeness hooks, forcing better results. Features like two-phase context compaction, persistent memory across sessions, and multi-LLM support (Anthropic, OpenAI, Gemini via litellm) make long research reliable without context blowups. Developers dig the GitHub integration vibes—easy setup with env vars for API keys like SERPER_API_KEY, and browser CDP mode for authenticated pages.

Who should use this?

AI researchers prototyping agent workflows, indie devs building internal knowledge tools, or analysts needing cited reports on tech trends without Perplexity subscriptions. Ideal for complex queries like policy comparisons or market overviews, where plain search falls short. Skip if you just want quick facts—it's overkill for one-offs.

Verdict

Promising early harness for deep agents (41 stars, 1.0% credibility), with solid docs and config-driven setup, but low maturity means watch for edge cases. Worth a spin locally if you need a controllable research agent—pip install, tweak settings.yaml, and run.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.