PaperPilot CLI
CLI agent for literature reviews: searches 11 open databases, screens papers and writes a bilingual cited report
- 167 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The program sends the topic and abstracts to the chosen language model and to public scholarly APIs
- Model and paid-source keys are stored in plain text in `~/.paperpilot/config.json`
- The agent downloads PDFs and writes files into the run folder on your disk
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/paperpilotDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The program sends the topic and abstracts to the chosen language model and to public scholarly APIs
- Model and paid-source keys are stored in plain text in `~/.paperpilot/config.json`
- The agent downloads PDFs and writes files into the run folder on your disk
This is third-party code. Review the repository files before installing.
What it does
PaperPilot takes a research request in plain language and turns it into a reproducible workflow. The agent first drafts a search protocol with inclusion and exclusion rules, then queries scholarly databases: arXiv, Semantic Scholar, OpenAlex, Crossref, OpenReview, PubMed, Europe PMC, bioRxiv and medRxiv, DBLP, ACL Anthology and Papers.cool. Results are normalized to one schema, duplicates are removed and papers are sorted into core, adjacent and excluded. Next it verifies links to PDFs and code, downloads open PDFs and optionally extracts their full text. For queries about code there is a `--github-filter` option: keep only works with a repository found, only works without one, or everything. Papers from your own folder or a BibTeX file are added with `--user-corpus`. The result is a report in Chinese and English as Markdown, HTML and PDF, with consistent citation numbering, a method table and an evidence matrix. Every run writes its state, event log and intermediate files into its own `runs/` folder, and `PaperPilot resume` picks up an interrupted job. An Obsidian note vault is generated as well.
Who it is for. Graduate students and researchers in AI, biomedicine and neighboring sciences who need a literature review with checkable references and a report in Chinese and English.
Good fit when
- You are entering a new topic and want a list of key papers with methods and code from a single run
- The review has to come out in two languages with matching citation numbers
- You have your own folder of PDFs or a BibTeX file to merge with open-database results
- You need a trail of the work: protocol, event log and evidence ledger for later checks
- You prefer to keep notes in Obsidian with links between papers, methods and topics
Not a fit when
- You need paywalled papers: PaperPilot does not bypass paywalls and downloads open PDFs only
- You want to run the search straight from Claude Code or Cursor: the tool is a standalone command-line agent with its own model profile
- A quick answer about one paper is enough: a full run with protocol and checks takes longer than an ordinary search
- You need clinical or legal expertise: the report gathers literature and does not replace a specialist's opinion
Example request
PaperPilot "RNA inverse folding sequence design" --auto-confirm --max-papers 50 --since-year 2021 --github-filter requiredLimitations
Needs a key for any OpenAI-compatible model: base URL, key and model name go into `~/.paperpilot/config.json`. Some sources, including CORE, Lens.org, IEEE Xplore, Springer Nature, Elsevier and Dimensions, switch on only with your own key. The online demo on Cloudflare Workers produces a lightweight metadata report, while PDFs, full text and the evidence ledger come only from the command-line version. The last commit in the repository dates from May 2026, and releases ship through PyPI.
How to disable. Run `python -m pip uninstall paperpilot`. Settings and keys live in `~/.paperpilot/config.json` and run results in `runs/` folders, which you delete by hand.
FAQ
How do I install PaperPilot?
Run `python -m pip install paperpilot`. On first launch it creates the template `~/.paperpilot/config.json`. You put the API address, key and model there or use `PaperPilot config set`, and `PaperPilot --doctor` checks the settings.
Which models work?
Any model reachable over the OpenAI protocol. The author's example uses DeepSeek with `deepseek-chat` and profiles switched by `PaperPilot config use`. The model handles request understanding, search planning, synthesis and the report text.
What is in a run folder?
`runs/<task-id>/` holds `task.json`, `state.json`, the event log, the search protocol, raw results, the screened corpus, downloaded PDFs, the evidence ledger, the paper matrix and the reports. Every step can be rechecked.
Can I skip the source API keys?
Yes. The free sources work by default: arXiv, OpenAlex, Crossref, PubMed and others. Keys are required only for CORE, Lens.org, IEEE, Springer, Elsevier, Dimensions and DeepXiv.
How do I continue an interrupted run?
`PaperPilot resume runs/<task-id>` restarts from the saved state, and `PaperPilot inspect runs/<task-id>` shows which step the process stopped at.
Discussion
No questions or reviews yet.
Related
CLI toolsFeynman — a terminal AI agent for scientific research: literature reviews, source checks, paper critique and comparing a paper with its code
Academic plugin for Claude Code, Gemini CLI, Codex and OpenCode: literature reviews, peer review and citation checks
A skill set that runs a research repository from its README within bounds and keeps verifiable evidence, without rewriting the original README
Self-hosted system for a team of AI agents: org chart, tasks, budgets and oversight of Claude Code, Codex and OpenClaw
An MCP server that splits math papers from arXiv, LaTeX and PDF into theorems, proofs and dependencies backed by the source
pwc CLI
pwc-cli
Hugging Face's CLI, skill and MCP server: an agent searches papers, benchmarks and top models in Papers with Code with no account or key
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| PaperPilot (this card) | no | 167 | ||
| Paper Search MCP | no | key is optional | 3,076 | an MCP server for searching and downloading papers from scholarly databases inside an agent chat |
| arxiv-mcp-server | no | not required | 3,273 | a local MCP server for arXiv only: reads papers by section and exports BibTeX |
| Claude Scholar | no | 6,036 | a Claude Code plugin with dozens of research skills rather than a standalone program |
README in short
The description opens with badges, online demo links and a feature overview: protocol, multi-source search, PDF and code checks, reports and an Obsidian wiki. Then come the source list, installation, model and key setup, launch examples, the process state diagram, output files and command-line options. Notes for contributors and the release routine close it.