Your coding agent already reads code, writes code, and runs experiments. ScholarAIO gives it a structured knowledge base of your research papers — so the same agent that writes your code can also search your literature, cross-check results against published findings, reproduce methods from papers, and draft your manuscript. One terminal, one agent, the full research loop.
Start Here
| If you want to... | Do this |
|---|---|
| Try ScholarAIO itself or contribute to the repo | Open this repository directly with your agent |
| Use ScholarAIO in Claude Code across many projects | Install the Claude Code plugin |
| Reuse ScholarAIO skills in Codex / OpenClaw | Register the skills through ~/.agents/skills/ |
Detailed setup guide: docs/getting-started/agent-setup.md
Use Inside This Repository
This is the best path when you want the full ScholarAIO experience: bundled agent instructions, local skills, CLI, and the complete codebase context.
# 1. Clone and install
git clone https://github.com/ZimoLiao/scholaraio.git
cd scholaraio
pip install -e ".[full]"
# 2. Configure your local environment
scholaraio setup
# 3. Start your agent in the repo root
claude
When you open the repo directly:
- Claude Code reads
CLAUDE.mdand.claude/skills/ - Codex / OpenClaw read
AGENTS.mdand.agents/skills/ - Cline reads
.clinerules - Cursor reads
.cursorrules - Windsurf reads
.windsurfrules - GitHub Copilot reads
.github/copilot-instructions.md
You can also use the CLI directly with scholaraio search "your topic".
Register ScholarAIO in Another Project
Claude Code plugin
ScholarAIO ships as a Claude Code plugin, so this is the cleanest cross-project install path:
Run these commands inside a Claude Code session, not in your system shell:
/plugin marketplace add ZimoLiao/scholaraio
/plugin install scholaraio@scholaraio-marketplace
After that, start a new Claude Code session in any project and use namespaced skills such as /scholaraio:search or /scholaraio:show.
Codex / OpenClaw skills
If you want ScholarAIO available to Codex-style agents outside this repo, clone it once and symlink the skills into the global discovery directory:
git clone https://github.com/ZimoLiao/scholaraio.git ~/.codex/scholaraio
cd ~/.codex/scholaraio
pip install -e ".[full]"
scholaraio setup
mkdir -p ~/.agents/skills
ln -s ~/.codex/scholaraio/.claude/skills ~/.agents/skills/scholaraio
Then make config discovery explicit for cross-project use:
# Option A: keep ScholarAIO data rooted in the cloned repo
export SCHOLARAIO_CONFIG="$HOME/.codex/scholaraio/config.yaml"
# Option B: move/copy config to the global fallback location
mkdir -p ~/.scholaraio
cp ~/.codex/scholaraio/config.yaml ~/.scholaraio/config.yaml
Without one of those two options, running scholaraio from another project may fall back to defaults rooted in that current project and create data/ plus workspace/ there. Restart the agent after creating the symlink. This registers the ScholarAIO skill library globally. For the full bundled project instructions, opening this repository directly is still the better path.