Article
AI agents are quickly gaining mastery over a growing share of work-critical tasks -- and doing for knowledge work what the engine once did for physical labor.
CopilotKit is bringing the capabilities of the entire LangChain & LangGraph chain & agent ecosystem, right into applications.
We'll first dive straight into the integration between CopilotKit and the LangChain / LangGraph ecosystems. We'll chat about the idea of human-in-the-loop agents and how they fit into software. To make things clear, we'll show you a Copilot-powered spreadsheet app that uses a research agent powered by Tavily and LangGraph. Lastly we'll end with a sneak peek at what's coming up for CopilotKit, including Co-Agents and self-learning features. Stick around!
We are quickly learning (and re-learning) that agents perform best when they work alongside people. It's far, far easier to get an agent to perform 70%, and even 80% and 90% of a given task, than to perform the given task fully autonomously (see: cars).
But to facilitate seamless human-AI collaboration, we must connect AI agents to the realtime application context: agents should be able to see what the end-user sees, and to do what the end-user can do in the context of applications.
And end-users should be able to monitor agents' executions - and to bring them back on track when they go off the rails. Simple enough - but where do you begin?
That's where CopilotKit comes in: CopilotKit is a framework and a platform for building AI assistants that are deeply integrated with applications. Think Cursor IDE (and its agentic heirs).. but for tax-planning software, insurance software, CRMs... and any other piece of software you can envision.