Agent Research Copilot (ARC) is a research toolkit for theoretical physics. It supports paper discovery and analysis, research-domain construction, proposer-reviewer loops, translation, companion readers, and source-aware calculation workflows.
ARC is designed for coding-agent hosts such as Codex, Claude Code, and similar agents.
Use ARC when you need to:
- acquire, parse, search, or summarize research papers;
- construct a source-aware research domain from seed literature;
- run proposer-reviewer idea or calculation loops;
- translate scientific sources or build chapter-aware companion readers; or
- use durable, inspectable research workflows instead of ad hoc prompts.
If ARC has played a role in your research, please consider citing the ARC manual.
Yanjiao Ma, Yi Wang, and Xingkai Zhang. ARC: An LLM-Native Agent Workflow for Theoretical Physics Research. ChinaXiv:202606.00234, 2026. https://chinaxiv.org/abs/202606.00234
@misc{ma2026arc,
title = {{ARC}: An {LLM}-Native Agent Workflow for Theoretical Physics Research},
author = {Ma, Yanjiao and Wang, Yi and Zhang, Xingkai},
year = {2026},
month = jun,
publisher = {ChinaXiv},
eprint = {202606.00234},
archivePrefix = {ChinaXiv},
url = {https://chinaxiv.org/abs/202606.00234},
note = {Version 1}
}-
Permission: the same as many heavy skills/plugins, ARC will need permissions to run Python scripts. Accepting permissions could be annoying. We recommend installing ARC within docker or a virtual machine, and allow all permissions in that virtual environment. As always for working with AI agents, be aware of risk to your data and system.
-
Token usage. As measured using Claude + DeepSeek, a typical run of domain build + idea generation consumes about 1M uncached input tokens, and 0.5M output tokens, in about an hour's running time. The token usage may vary depending on the specific tasks and LLM used. Be aware of token usage and costs.
codex plugin marketplace add tririver/arc --ref stable
codex plugin add arc@arc/plugin marketplace add tririver/arc@stable
/plugin install arc
Give your coding agent this repository and ask it to inspect the repository and install ARC for its environment.
After installing ARC, ask for the research outcome directly:
Use ARC to summarize a paper.
Use ARC to build a domain from arXiv:0911.3380 with new papers since 2024.
Use ARC to develop and review ideas from the resulting domain.
Use ARC to check this calculation.
The ARC manuals provide concise quick starts. Use built-in --help for exact
commands, options, and error guidance.
ARC development requires Python 3.11 or newer. Read AGENTS.md before making
changes.
Keep research runs and generated output below the git-ignored local/ tree.
Run focused package tests first, then the combined offline test and build checks:
python -m pytest --import-mode=importlib packages/*/tests
scripts/check-packages.shNetwork and live-model tests are opt-in. Do not use them as the default development check.
Releases are explicit human operations from a clean release checkout:
scripts/release-arc.sh <version>The helper validates the release, updates package and plugin versions, and
pauses before its mutating Git steps. See AGENTS.md for repository
development, verification, and release constraints.