AI tools and harnesses
Assistants, agents, coding environments (IDEs), notebooks, and platforms that wrap AI models into something you actually use to get work done. If a model is the engine, a harness is the car built around it. Ordered alphabetically by title.
Coding Harnesses
Codex
Level: Intermediate · Modality: Code · Purpose: Coding, Productivity, Research
OpenAI’s coding agent. Fellows used it as a local coding assistant that can read a codebase, make changes, and iterate, rather than a plain chat window.
Reviews:
- 07/2026 — Dezhe Jin — Recommend. Especially useful for turning research code into more reliable, professionally structured software.
- 07/2026 — Chaoxing Liu — Recommend. Used alongside Claude-style coding agents for materials-simulation work, including step-by-step verification of derivations and reproduction of published figures and models; extending beyond existing published results still required substantial expert effort.
Source: 2026 Summer Faculty Upskilling Reflection Lunch meeting notes.
Claude Code
Level: Intermediate · Modality: Code · Purpose: Coding, Productivity, Research
Anthropic’s terminal-based coding harness. Fellows used it as a local coding assistant that can read a codebase, make changes, and iterate, rather than a plain chat window.
Reviews:
- 07/2026 — Eric Ford — Highly Recommend. Excellent for understanding repositories, translating codes across languages, writing documentation, tests, setting up continuous integration.
Source: 2026 Summer Faculty Upskilling Reflection Lunch meeting notes.
Penn State AI Studio
Level: Beginner · Modality: Text · Purpose: Productivity, Research
Penn State’s own AI assistant environment, provided for the university community. A safe first place to experiment with a chat assistant using university credentials.
Reviews:
- 07/2026 — Penn State — Recommend. A sensible starting point for faculty who want to try an assistant within the university’s supported environment rather than a personal account. Approved for Level 1 and 2 Data.
PDF-to-Markdown converters
Level: Beginner · Modality: Text, Code · Purpose: Research, Productivity
Converting PDFs to Markdown makes papers and reports easier to search, edit, and feed into AI tools. Several converters are worth trying rather than just one.
Reviews:
- 07/2026 — Sona Golder — Highly Recommend. See the detail page for specific tools tried.
- 07/2026 — Eric Ford — Highly Recommend. See the detail page for specific tools tried, including markitdown and arxiv2md.
More info: PDF-to-Markdown converters (detail page)
Token saving strategies
RTK
Level: Intermediate, Advanced · Modality: Text, Code · Purpose: Research, Productivity
Can significantly reduce token usage.
Reviews:
- 07/2026 - Justin Silverman - Highly Recommend. Helps keep usage within limits of low cost subscription tiers
To add an entry, copy the block in templates/resource-entry-template.md. See CONTRIBUTING.md.
As with models, confirm a tool is approved for your data before using it for university work. See Penn State’s AI Services and Best Practices for Using AI Tools.