"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis Your Agent's Self-Improving Swiss Army Knife: Composio CTO Karan Vaidya on Building Smart Tools
126 snips
Mar 22, 2026 
Karan Vaidya, CTO of Composio and builder of AI agent infrastructure, maps out a smart tool layer for agents. They dig into tool discovery, auth, sandboxes, and self-healing workflows. The conversation also explores reducing model lock-in, translating skills across providers, safer permission profiles, and why the biggest wins come from agents tackling full jobs.
AI Snips
Chapters
Transcript
Episode notes
Smart MCPs Need Progressive Disclosure
- A smart MCP should reveal tools progressively, not dump thousands into context and hope the model sorts them out.
- Composio adds just in time tools, rewrites broken tools in real time, and converts successful trajectories into reusable skills plus failure warnings.
A Hiring Agent Already Does Recruiter Work
- Karan Vaidya uses an OpenClaw hiring agent to run a full recruiting workflow, not just isolated tasks.
- It scans GitHub commits in agentic Python and TypeScript repos, enriches candidate data, emails prospects, and booked roughly 30 to 40 calls in one to two weeks.
Agent Context Should Be Split By Permission Profile
- The future agent stack likely uses multiple agents with different permission profiles rather than one all powerful assistant.
- Karan Vaidya suggests read only research agents with broad data access, and separate write enabled agents with tighter context plus human approval.