Frequently asked questions

A chatbot answers; an agent acts. Agents use the model to decide their next step — calling tools, reading data, checking their own work — in a loop, until the goal is done. The chat interface may look identical; the loop underneath is the difference. Build the loop in 30 minutes →

No — the core agent loop is about 40 lines of code, and for one agent with a handful of tools, rolling your own is less complexity, not more. Frameworks (LangGraph, CrewAI, the OpenAI Agents SDK, Microsoft Agent Framework, Pydantic AI, Google ADK and others) earn their keep at durable long-running state and parallel multi-agent graphs. Build the bare loop once first, then decide. The build-vs-buy verdict →

No. MCP is an open standard hosted by the vendor-neutral Agentic AI Foundation, with clients across the ecosystem: Claude Code and Codex, editors such as VS Code and Cursor, agent frameworks, and even llama.cpp's local server. A server you write once works in any compliant client. Build your first server →

The stable answer is a method, not a model name: route by task tier — a frontier model (Claude, GPT, Gemini class) for planning, cheaper or open-weights models for routine steps — and let golden-task evals on your own workload pick the names. Leaderboards measure chat, not multi-step tool reliability. The six criteria that matter →

Yes — open-weights models served by llama.cpp or vLLM expose an OpenAI-compatible endpoint, and the rest of the agent stack (harness, MCP tools, evals) is just your code. llama.cpp's server can even launch local MCP servers itself (an experimental feature). Expect small local models (roughly 7–14B) to be weaker at long-horizon planning than frontier hosted models; measure on your tasks. Local serving guide →

They can be — if safety comes from architecture rather than prompts. The controls that work: capability allowlists at the tool layer, approval gates for irreversible actions, trajectory logging, and evals run on every change. "We told the model to be careful" is not a control. The working threat model →

Agent cost = usage × verbosity × retries × context length — four multipliers one commit can change. Meter per trajectory (not per call), route task tiers to appropriately-priced models, and cap retries in the harness. Instrument before you scale; dashboards of regrets are expensive. FinOps for agents →

Test outcomes and trajectories, not phrasing: golden tasks with checks like "the answer contains the right fact", "the right tool was called", "the forbidden tool was not". Run tasks multiple times, gate CI on pass rates, and turn every production incident into a new golden task. Build the eval harness →

It is malicious instructions hidden in content your agent reads — a web page, an email, a document. If your agent has tools and touches untrusted content, yes: assume injection sometimes succeeds and design so a successful one cannot reach anything worth the attack. OWASP published a peer-reviewed Top 10 for Agentic Applications in December 2025; use it as your checklist. Defenses that actually work →

Use the Start Here page — pick a goal and get a numbered path. If you just want the fastest meaningful win: build the 40-line agent loop, then your first MCP server. Both run on a laptop with no API key; only connecting the server to a chat app needs an account. Start Here →

Pick your route. Start Here asks what you want to do and gives you a short path. Learn lays every guide out as a track, from first agent to production, and each article links to the next step at the bottom. Tutorials are hands-on, Compare pieces help you decide, Enterprise covers governance, the Radar is the quarterly view of what to adopt, and Templates are copy-paste starters. See the learning tracks →

It helps, but no. The tutorials use Python because it is the shortest to read, and the key ones (like the MCP server) also have a TypeScript version. If you can read a loop and a function, you can follow along. Coding agents such as Claude Code or Codex can explain any line you get stuck on. Try the 40-line agent loop →

For new servers, use the v2 SDKs: the mcp 2.x package in Python (the old FastMCP class is now MCPServer) or @modelcontextprotocol/server in TypeScript. They support the stateless 2026-07-28 spec revision and still talk to older clients, many of which have not upgraded yet. If an existing tutorial or project imports mcp.server.fastmcp, either migrate it or pin mcp below version 2. Migrate to the 2026 spec →