skip to content
superbot.gg

Search papers and GitHub repos from an agent

Your agent searches scholarly papers.
It searches GitHub repositories too.

Get the app see the real call

What you get

Ask your agent

Connect superbot once.
Other clients are on the install page.

claude mcp add --transport http superbot https://api.superbot.gg/mcp -s user

With no browser, post to the install route instead.

curl -fsS -X POST https://api.superbot.gg/mcp/install -d client=claude-code

Then ask in plain words.

superbot({"intent": "Find papers on speculative decoding and GitHub repos that implement it."})

This needs an account.
An agent with none signs you in through /auth.md.
You read back one emailed code.

One real call

A real call, run on 2026-10-18 from a test account.
The reply is the agent's own, with its heading, its sources block and an account label cut.
Em dashes are shown as hyphens.

call
superbot({"intent": "Find three scholarly papers on retrieval augmented generation and three GitHub repositories for it, with links."})
reply
Three RAG papers and three repos.

**Papers** (from OpenAlex, read 2026-10-18):

1. [**Retrieval-Augmented Generation for Large Language Models: A Survey**](https://arxiv.org/pdf/2312.10997) - Gao et al., 2023. The standard survey: naive, advanced and modular RAG, cited 871 times.
2. [**Active Retrieval Augmented Generation**](https://aclanthology.org/2023.emnlp-main.495.pdf) - Jiang et al., EMNLP 2023. Retrieval decided while generating (FLARE), cited 472 times.
3. [**Benchmarking Large Language Models in Retrieval-Augmented Generation**](https://ojs.aaai.org/index.php/AAAI/article/download/29728/31250) - Chen et al., AAAI 2024. Recite, Noise Robustness, Counterfactual Robustness and Information Integration benchmarks, cited 444 times.

**Repositories** (from GitHub, read 2026-10-18):

1. [**infiniflow/ragflow**](https://github.com/infiniflow/ragflow) - 92,401 stars, Apache-2.0, pushed 2026-10-17. Open-source RAG engine with agent features.
2. [**HKUDS/LightRAG**](https://github.com/HKUDS/LightRAG) - 40,198 stars, MIT, pushed 2026-10-15. Graph-based RAG, the code for the EMNLP 2025 LightRAG paper.
3. [**microsoft/graphrag**](https://github.com/microsoft/graphrag) - 36,370 stars, MIT, pushed 2026-10-12. Modular graph-based RAG system.
A paper and repo search, run live. Three of each.

Limits

Needs
A superbot account.
As of
Counts are those of the day of the search. In the call above, stars and citations are the figures on 2026-10-18.

Papers and repositories as they stood on the day of the call. Read on 2026-10-18.