Skip to main content
github.com/onecortex-io/examples holds one agent per folder, one per supported framework. None of them needs a model API key: each uses a scripted model and the data in its folder, so every answer is exact, while the events it streams are the ones its framework really produces. Each folder’s README says what to ask it and where a real model plugs in. MIT licensed.

​
Deploy one

  1. Fork the repository.
  2. In the dashboard, click Create agent and connect GitHub. See Connect GitHub.
  3. Pick your fork, and choose a folder as the Agent folder.
  4. Click Create agent, and call it once the build succeeds. See the invoke API.
The quickstart walks through this with echo.

​
Python

echo

A plain function: one tool call of its own, then You said: <prompt>. The quickstart’s agent.

support-triage

LangGraph: classifies a ticket, looks up the order, remembers the session.

travel-concierge

OpenAI Agents SDK: triage hands off to booking, which searches flights.

weather-planner

Strands Agents: checks the forecast, then plans the day.

docs-qa

LlamaIndex: searches router docs and cites the page.

expense-assistant

LangChain AgentExecutor: converts and records expenses.

research-crew

CrewAI: a researcher and a writer produce a brief.

​
TypeScript

recipe-assistant

Mastra: finds recipes, honouring a diet the caller sends.

meeting-scheduler

Vercel AI SDK: checks the calendar and holds a slot.

ticket-summarizer

LangChain.js: classifies a ticket and looks up the customer.

order-status

A plain function: looks up an order with its own tool events.
Last modified on September 28, 2026