Coding with AI
How Tools Like Cursor Are Changing the Developer's Workflow
https://slides.com/agesteira/coding-with-ai-cursor
-> The Chat
-> The Landscape
-> The IDE
-> The Workflow
-> The Review
Andrés Gesteira
/in/andres-gesteira-26a95813
Born in Madrid
Senior Frontend Architect @ Fever
NG Poland lover
Likes cinema, books, and food...
@GeorgeKaplan_G
LLM Chat vs Agent Chat
- LLM:
- Sends a message and the model replies.
- Everything runs synchronously.
- No autonomy. The user always decides.
- Agent:
- Has a goals and autonomy. So it can make decisions.
- Takes actions in one or several steps. Each action in a loop.
- Perception → Reasoning → Action → Review → Repeat
- Has access to tools (MCPs, system terminal, etc).
Too many agents!
So let us go through them quickly
GitHub Copilot
Cline & Roo Code
JetBrains AI Assistant
Windsurf
Kiro
Trae
Codex
Claude Code
Warp
Antigravity
Cursor
The Cursor Agent
Previously known as "Composer". It is a single set of context and requests, which could involve multiple returns to an LLM, potentially involving different LLMs or small Cursor models for context fetching, often constituting one sequential chat.
The Cursor Environments
- Local (Agent, Plan or Ask)
- Either in your branch
- ...or in a Git Worktree (sandboxed clone). Requires setup configurations (`worktrees.json`).
- Cloud:
- Long running in a virtual machine.
- Full autonomy (branches, PRs etc) no MCPs or web search.
- Can be integrated with tools like Slack.
The Cursor Context
- Prompts
- System
- Rules (persistent instructions).
- User input
- Project: all files except:
- `.gititnore`
- `.cursorignore`
- Special extensions e.g. `.env`.
- MCP output
The Cursor Rules
- User (prio1)
- Project (prio2)
- Team (prio3)
- `AGENTS.md`. Treated as project rules for:
- Folder structure
- Technology Stack
- Ownership
- Workflows and workflow conditions e.g. MCP tools vs CLI commands.
The Cursor Data Flow
Spec Driven Development
The Study
The Task
The Plan
The Implementation
The Documentation
Code Review + Cursor Rules
- `.cursor/rules/**/*.mdc`
- `.cursor/BUGBOT.md`
- `.github/copilot-instructions.md`
- name: Perform code review
env:
CURSOR_API_KEY: ${{ secrets.CURSOR_API_KEY }}
GH_TOKEN: ${{ github.token }}
run: |
cursor-agent --force --model "$MODEL" --output-format=text --print "You are operating
in a GitHub Actions runner performing automated code review. The gh CLI is available and
authenticated via GH_TOKEN. You may comment on pull requests.
Context:
- Repo: ${{ github.repository }}
- PR Number: ${{ github.event.pull_request.number }}
- PR Head SHA: ${{ github.event.pull_request.head.sha }}
- PR Base SHA: ${{ github.event.pull_request.base.sha }}
Objectives:
1) Re-check existing review comments and reply resolved when addressed
2) Review the current PR diff and flag only clear, high-severity issues
3) Leave very short inline comments (1-2 sentences) on changed lines only and a brief summary at the end
Procedure:
- Get existing comments: gh pr view --json comments
- Get diff: gh pr diff
- If a previously reported issue appears fixed by nearby changes, reply: ✅ This issue appears to be
resolved by the recent changes
- Avoid duplicates: skip if similar feedback already exists on or near the same lines
Commenting rules:
- Max 10 inline comments total; prioritize the most critical issues
- One issue per comment; place on the exact changed line
- Natural tone, specific and actionable; do not mention automated or high-confidence
- Use emojis: 🚨 Critical 🔒 Security ⚡ Performance ⚠️ Logic ✅ Resolved ✨ Improvement
Submission:
- Submit one review containing inline comments plus a concise summary
- Use only: gh pr review --comment
- Do not use: gh pr review --approve or --request-changes"
Thanks
