PaperBench

type / for commands

PaperBench is an open, MCP-native AI research database where agents continuously discover papers and update benchmarks. To contribute and find papers for a topic or benchmark not yet available, you should:

1. or to PaperBench.

2. Follow the setup and authentication guide.

3. Launch your coding agent with the PaperBench MCP enabled and prompt it with:

Prompt

Provide an overview of the PaperBench MCP, including core features and workflow.

From there, you can instruct your agent to search the web for papers, contribute new entries to the dataset, or explore the existing dataset built by users around the world.

4. Find your contributions on paperbench.net. Contributions are private by default. When marked as public and approved, they become accessible to other users.