Practical LLM Apps

Clone-and-run AI agents, RAG apps, and MCP servers.

Hand-built examples you can customize and ship — not slide-deck demos.

Works with OpenAI, and several apps run fully local with no API key.

Run one now

Invoice Follow-up Agent

git clone https://github.com/nimish14/practical-llm-apps.git
cd practical-llm-apps/starter_ai_agents/invoice_followup_agent
pip install -r requirements.txt
cp .env.example .env   # add OPENAI_API_KEY
streamlit run app.py

Chat with PDF

cd practical-llm-apps/starter_ai_agents/chat_with_pdf
pip install -r requirements.txt
cp .env.example .env   # add OPENAI_API_KEY
streamlit run app.py

Resume Evidence RAG (no API key)

cd practical-llm-apps/rag_apps/resume_evidence_rag
pip install -r requirements.txt
streamlit run app.py

Freelance Ops MCP (Cursor / Claude)

cd practical-llm-apps/mcp_servers/freelance-ops-mcp
npm install
npm run build
npx -y @modelcontextprotocol/inspector node dist/index.js

Add to Cursor MCP settings:

{
  "mcpServers": {
    "freelance-ops": {
      "command": "npx",
      "args": ["-y", "freelance-ops-mcp"]
    }
  }
}

(After npm publish — until then, point args at node + local dist/index.js.)

Browse templates

Starter AI Agents

App	Description
Invoice Follow-up Agent	Draft polite → firm invoice reminders with tone control and overdue context
Chat with PDF	Ask questions about an uploaded PDF with page citations

RAG Apps

App	Description
Resume Evidence RAG	Hybrid MiniLM + BM25 retrieval that returns grounded resume facts for a job description

MCP Servers

Server	Description
freelance-ops-mcp	Tools for overdue status, tone suggestion, and invoice reminder drafts
resume-evidence-mcp	BM25 resume evidence retrieval tools for coding agents

Project goals

Every app: clone → install → run in under 2 minutes
Clear README with problem, pipeline, and limits
MCP servers packaged for npm + MCP Registry
Prefer small, focused tools over mega-frameworks

License

MIT — fork it, ship it, sell it.

practical llm apps