Want to build AI agents without incurring cloud API costs? Discover how to run a private, local AI agent using Ollama, LangGraph, and MCP. Running generative AI (gen AI) agents on third-party cloud infrastructure often leads to unpredictable token costs and data privacy concerns. Running open source models locally provides a private, cost-effective alternative for development. In this video, learn how to serve an open-weight Qwen3 large language model (LLM) on your local machine using Ollama and ogx-server. Explore how to orchestrate agent workflows with LangGraph and attach 3 distinct tools: a Python calculator, a web search application programming interface (API) via Firecrawl, and a GitHub integration using an external MCP server. 00:00 Opening: Token costs are up 00:10 The case for local AI development 00:45 Let's build: Agent, chat UI, and custom tools 01:58 Serving Qwen3 with Ollama and ogx-server 02:48 Building the agent loop in Python 03:35 Registering tools: Calculator, web search, and GitHub MCP 04:50 Local playground demo 05:29 Wrap-up and repository instructions Explore more open source AI tools and resources: 📂 Access the AgentOps GitHub repository → https://red.ht/github-agent-ops 🤖 Check out Red Hat AI's Hugging Face → https://red.ht/rhai-hugging-face 💡 Learn more about AI inference → https://www.redhat.com/en/topics/ai/what-is-ai-inference?sc_cid=RHCTG0260000496899 #MCP #Ollama #LangGraph #GenerativeAI #OpenSource #RedHat

 19724      313