AI Research Workflows
Guides for integrating Claude Code into academic research. From configuring context files to building a personal AI infrastructure that improves over time.
AI-Assisted Research: What Changes and What Stays the Same
The integration of AI into research workflows changes the speed of implementation, not the standard of evidence. A well-configured Claude Code session can translate a methodology paragraph into working code in minutes. It can refactor a sprawling research codebase into something maintainable. It can draft documentation while the analysis is still fresh.
What it cannot do is replace the researcher's judgment about what to measure, how to interpret results, or when to be skeptical of convenient findings. The domain expertise that takes years to develop remains essential. AI accelerates the translation of that expertise into artifacts: code, documentation, analysis pipelines.
The key artifact in this workflow is the CLAUDE.md file. It loads automatically when Claude Code starts a session, providing project requirements, variable definitions, data constraints, and methodological decisions in one place. Without it, every session starts from zero. With it, the AI reads the same context the researcher would provide to a new collaborator, and the conversation picks up where it left off.
These guides document the practices we developed over hundreds of research sessions. The verification tax, the cold start problem, context window budgeting, end-of-session hygiene: each addresses a specific failure mode that surfaces when AI assistance is integrated into work that demands accuracy. The goal is not to automate research, but to make the human-AI collaboration more reliable.
[1] Anthropic. "Claude Code overview." docs.anthropic.com/en/docs/claude-code/overview, accessed February 2026.
[2] Anthropic. "Claude Code best practices: CLAUDE.md." docs.anthropic.com/en/docs/claude-code/claude-md, accessed February 2026.
Agent-based vs. chat-based AI for research
| Capability | Claude Code (Agent) | Chat-based AI |
|---|---|---|
| Reads project files | Directly, across full codebase | Only what is pasted in |
| Writes and edits files | Yes, in place | Copy-paste only |
| Persistent context | CLAUDE.md loads automatically [2] | Manual re-explanation each session |
| Runs shell commands | Yes (git, python, tests) | No |
| External data sources | MCP servers (FRED, Census, Scholar) | Web search only |
| Custom automation | Skills, hooks, sub-agents | None |
| Cost | $20/month (Pro) or usage-based [1] | $20/month (typical) |
Getting Started
Your First Session: What Claude Code Is and Isn't
A practical walkthrough of what Claude Code can and cannot do, with prompting patterns and a complete first-task example.
Claude Code Guide: From First Session to Personal AI Infrastructure
A comprehensive 13-article guide to mastering Claude Code. Context management, session workflows, agent spawning, and building personal AI infrastructure.
Why It Forgot Everything: Understanding Context
Understanding how AI context windows work, why sessions reset, and how to work with this fundamental limitation of large language models.
Claude Code Guide Series
Creating Skills for Research
Skills are recipe cards for research tasks. Write the steps once, save them in a file, and Claude Code follows those instructions whenever needed.
Creating Helpers: When to Delegate Work
When to create separate Claude Code helpers for focused work, how to design tasks that are easy to hand off, and patterns for running multiple helpers at once.
Building Our Research System: Putting It All Together
How CLAUDE.md, skills, hooks, and MCP servers combine into a personal research system that becomes more valuable over time.
Hooks: Automation Without Asking
Hooks are automatic triggers that run without asking, like auto-save but for research tasks. A power-user feature, entirely optional.
Connecting Claude to Outside Services: FRED, Census, and Beyond
How to connect Claude Code to external data sources like FRED, Census, and Google Scholar, bringing integrated research workflows into natural conversation.
Research Workflow
One Context File, Zero Re-Explanations
Stop re-explaining the project every AI session. A single CLAUDE.md file loads automatically. Write context once; the agent reads it every time.
From Methods Paragraph to Working Pipeline
A well-written methodology section is almost executable code. The gap between describing a procedure and implementing it has narrowed with agent-based tools.
Research Phases Need Different Prompts
Exploration, implementation, and documentation require different AI prompting strategies. Match the prompt to the phase.
Question-First Data Tagging: Finding Forgotten Datasets
Tag data by the questions it can answer, not just what it contains. How question-first tagging turns dormant datasets into discoverable assets.
Code Management
47 Scripts to 15: Cleaning a Research Codebase
Research codebases accumulate cruft. We used Claude to consolidate 47 scripts to 15, with counterfactual tests proving nothing broke.
Reading Our Analysis Files: How Claude Sees Our Research Code
How Claude Code explores research projects using three core tools: Read (look at a file), Glob (find files by pattern), and Grep (search inside files).
The Limits of Copy-Paste AI Coding
The difference between chatbot-based and agent-based coding is categorical, not incremental. What changes when AI reads the entire codebase.
The Verification Tax: Every AI Output Needs Checking
Every AI output needs checking. Building verification into workflow to catch hallucinations before they compound.
Advanced Topics
Context Window Budgeting: Treating Tokens as a Finite Resource
Treating tokens as a finite resource, and knowing when to spawn agents versus work directly.
The Cold Start Problem: Why the First Five Minutes Matter Most
Why the first five minutes of an AI session matter most, and how CLAUDE.md solves the context problem.
End-of-Session Hygiene: What to Capture Before Context Resets
What to capture before context resets, and how five minutes of capture saves twenty minutes tomorrow.
Monitoring Government Data Portals
A case study in tracking California health data releases with Claude Code. Catch new data releases without manual checking.
Building a Literature Surveillance Skill
Automating academic paper discovery with Claude Code. Turn weekly manual searches across SSRN, NBER, and Google Scholar into a single command.
Staging LinkedIn Posts with Browser Automation
A case study in form-filling workflows that keep humans in the loop. Browser automation handles navigation while the human retains final approval.
Fix AI Data Visualization: Why Claude Fails (+ Solution)
AI writes matplotlib code but cannot see if labels overlap or legends clip. Antigravity prompts solve this.
Reading Your Own Data: What Claude Code /insights Reveals
How to interpret the Claude Code /insights report at beginner and intermediate levels. Same data, different lessons.
Key takeaways
- AI changes speed, not standards. A well-configured Claude Code session translates methodology into working code in minutes, but domain expertise still determines what to build.
- CLAUDE.md eliminates the cold start problem. One context file loads project requirements, variable definitions, and methodological decisions automatically each session.
- Every AI output needs checking. The verification tax is real: code compiles and runs but may contain subtle errors that compound across a research project.
- 24 guides from hundreds of sessions. Each addresses a specific failure mode: context resets, hallucinated citations, token budget exhaustion, and more.
Frequently Asked Questions
What is Claude Code?
Claude Code is a command-line AI tool from Anthropic that reads and writes files in a project directory. Unlike chat-based AI, it has direct access to the codebase and maintains context through CLAUDE.md files.
Do I need programming experience?
Basic command-line familiarity is helpful. The guides focus on configuring Claude Code through context files and workflow design, not on writing code from scratch.
What is the verification tax?
The time and effort required to check every AI output. AI-generated code compiles and runs but may contain subtle errors. Our verification tax article explains how to build checking into the workflow.
Can AI replace domain expertise?
No. AI accelerates the translation of expertise into code, documentation, and analysis. The researcher still needs to understand what to measure, how to interpret results, and when to be skeptical.