AI Agents Projects

Beginner Projects

1. Calculator Agent

Build an agent with add, subtract, multiply, divide, and sqrt tools. Have it solve multi-step math word problems by calling tools in sequence and showing its reasoning.

What you’ll practice: Basic tool definition, tool call parsing, multi-step reasoning

2. File System Assistant

Build an agent with read_file, write_file, list_directory, and search_file tools. Have it organize a messy folder, rename files by convention, and generate a folder summary report.

What you’ll practice: File system tools, safety checks (no deletes without confirmation), multi-step planning

3. Web Research Assistant

Build an agent with a web_search tool (Serper API or similar). Have it research a topic, synthesize findings from 5+ sources, and produce a structured summary with citations.

What you’ll practice: Search tool integration, information synthesis, citation tracking

4. Weather Reporter Agent

Build an agent that calls a weather API, interprets the data, and generates a friendly human-readable weather report. Handle API errors gracefully and ask for location clarification when ambiguous.

What you’ll practice: API tool integration, error handling, natural language generation from data

5. SQL Query Agent

Build an agent with a run_sql tool (read-only). Given a database schema, have it answer natural language questions by generating and executing SQL, explaining results, and handling query errors.

What you’ll practice: Tool-generated SQL, error feedback loop, schema injection

6. Unit Conversion Agent

Build an agent that handles any unit conversion (length, weight, temperature, currency, time zones). Use tools for live currency rates; compute others directly. Handle ambiguous conversions by asking.

What you’ll practice: Tool routing, live data integration, ambiguity resolution

7. Shopping List Optimizer

Build an agent that takes a recipe list, identifies common ingredients, finds the cheapest combination, and produces a consolidated shopping list sorted by store aisle.

What you’ll practice: Multi-step reasoning, list manipulation tools, optimization thinking

8. Git History Analyzer

Build an agent with git_log, git_diff, and git_blame tools. Have it answer questions like “what changed most in the last month” and “who owns this file” and “summarize changes to the auth module.”

What you’ll practice: CLI tool wrappers, code analysis, technical reasoning

9. Data Cleaning Agent

Build an agent that inspects a CSV, identifies data quality issues (nulls, outliers, duplicates, format inconsistencies), proposes a cleaning plan, and executes it step by step with user approval.

What you’ll practice: Human-in-the-loop, data inspection tools, approval workflow

10. Meeting Scheduler Agent

Build an agent with check_calendar, find_free_slots, and send_invite tools. Given attendee names and constraints (“Tuesday afternoon, 1 hour”), find the best slot and draft the invite.

What you’ll practice: Tool chaining, constraint satisfaction, calendar API integration

Intermediate Projects

1. Autonomous Code Debugger

Build a debugging agent: read code + error message, hypothesize causes, modify code (in a sandbox), run tests, check if fixed, and iterate up to 5 rounds. Track fix rate and round count.

What you’ll practice: Code execution tools, self-correction loop, iteration budget

2. Competitive Analysis Agent

Build an agent that: takes a company name, searches for competitors, finds pricing pages, extracts features/prices, generates a comparison table, and identifies market gaps.

What you’ll practice: Multi-step web research, structured extraction, competitive intelligence

3. Legal Document Reviewer

Build an agent that reviews contracts clause by clause: identifies risky language, flags missing standard clauses, suggests alternatives, and produces a summary risk report with severity ratings.

What you’ll practice: Document analysis tools, structured output, domain prompting

4. Personal Finance Advisor Agent

Build an agent with tools to read transaction CSVs, categorize spending, compute budget vs. actual, identify savings opportunities, and generate a monthly financial health report.

What you’ll practice: Data analysis tools, financial reasoning, report generation

5. Multi-Source News Synthesizer

Build an agent that fetches the same story from 5+ sources, compares coverage (what’s included/omitted in each), identifies potential bias, and produces a balanced synthesis.

What you’ll practice: Parallel tool calls, bias detection prompting, synthesis

6. Automated PR Reviewer

Build a GitHub agent: fetch PR diff, analyze each file for bugs/style/security, check if tests cover new code, verify documentation is updated, and post a structured review comment.

What you’ll practice: GitHub API tools, code review reasoning, multi-file analysis

7. Resume Tailoring Agent

Build an agent that reads a job description and resume, identifies gaps, suggests specific changes (not just generic advice), rewrites weak bullet points, and scores the final match.

What you’ll practice: Document comparison, targeted suggestion generation, iterative refinement

8. Knowledge Base Builder

Build an agent that: reads a domain corpus, extracts key concepts and relationships, stores them in a structured knowledge base, and answers questions by traversing the knowledge graph.

What you’ll practice: Extraction tools, knowledge storage, graph traversal, Q&A

9. Incident Response Agent

Build an agent for on-call engineers: reads alert + logs, hypothesizes root causes, queries monitoring tools, suggests remediation steps, and drafts an incident report — all in under 2 minutes.

What you’ll practice: Time-bounded agents, monitoring tool integration, technical synthesis

10. Content Marketing Agent

Build an agent that: given a topic and audience, researches the web, generates a content brief, writes a draft, checks factual claims, optimizes for SEO, and produces a final polished post.

What you’ll practice: Multi-step content pipeline, fact-checking tools, SEO integration

Advanced Projects

1. Self-Improving Agent

Build an agent that: runs a task, evaluates its own output, identifies failure patterns, generates improved strategies, and tracks performance improvement over N iterations.

What you’ll practice: Self-evaluation, meta-learning loop, performance tracking

2. Multi-Agent Debate System

Build a multi-agent debate: two agents argue opposite positions, a moderator asks clarifying questions, a judge scores arguments on logic and evidence, and a synthesizer produces the best combined view.

What you’ll practice: Agent roles, adversarial reasoning, structured debate protocols

3. Autonomous Data Scientist

Build a data science agent: accepts a dataset and question, performs EDA, selects features, tries 5 models, tunes the best one, evaluates honestly, and generates a findings report with code.

What you’ll practice: End-to-end automation, code execution tools, ML reasoning

4. Long-Horizon Planning Agent

Build an agent for a 10-step task with dependencies: create a DAG of sub-tasks, identify which can run in parallel, checkpoint progress to disk, recover from partial failures, and complete the goal.

What you’ll practice: Task graphs, parallel execution, checkpointing, fault tolerance

5. Agent Safety Evaluator

Build an evaluation harness for agents: 100+ test cases covering prompt injection, jailbreak attempts, scope creep, infinite loop potential, and unsafe tool use. Score agents on safety + capability tradeoffs.

What you’ll practice: Red-teaming, safety evaluation, test design, scoring methodology

Portfolio Projects

1. Autonomous Software Engineer Agent

Build an agent that accepts GitHub issues, creates a branch, writes code fixes, adds tests, verifies tests pass, and opens a PR — end to end. Handle ambiguous issues by asking clarifying questions.

Tech stack: Anthropic API, GitHub API, Docker sandbox for code execution
Demonstrates: Autonomous coding, system integration, safety (sandboxed execution)

2. Enterprise Research Intelligence Platform

Build a multi-agent research system: specialized agents for web search, document analysis, data extraction, fact verification, and report writing. Produce verified, cited research reports at scale.

Tech stack: Anthropic API, multi-agent framework, vector database, FastAPI
Demonstrates: Multi-agent orchestration, production scale, research quality

3. AI-Powered DevOps Assistant

Build a DevOps agent with tools for: reading logs, querying metrics, checking deployments, running diagnostics, and executing remediation scripts. Handle on-call alerts autonomously for known incident patterns.

Tech stack: Anthropic API, Kubernetes API, Prometheus API, PagerDuty
Demonstrates: Production systems knowledge, autonomous operations, safety controls

4. Personalized Learning Coach

Build a learning agent that: assesses a learner’s current knowledge, generates a personalized curriculum, creates exercises at the right difficulty, gives explanations, tracks progress, and adapts based on performance.

Tech stack: Anthropic API, SQLite for progress tracking, Streamlit UI
Demonstrates: Adaptive AI, long-term memory, personalization, educational design

5. Autonomous Financial Analyst

Build an agent that: reads earnings reports and filings (SEC EDGAR), extracts key metrics, computes financial ratios, compares to historical trends and peers, and produces an investment research note with risks and opportunities.

Tech stack: Anthropic API, SEC EDGAR API, financial data APIs, PDF parsing tools
Demonstrates: Domain expertise, financial reasoning, research automation