← back
Just do it. (let your tools think for themselves) - Robert Chandler
Takeaway
Make MCP tools agentic — give them natural-language inputs and let them reason internally — instead of exposing thin API wrappers that overflow the parent agent's context.
Summary
- Robert Chandler (Wordware CTO, ex-self-driving) argues most MCPs are low-level API wrappers that pollute context with JSON blobs and force agents to chain many tiny function calls.
- Memorable failure: a Slack MCP took 5 minutes and ended up posting 'could someone tell Philip MCP is awesome?' in #general after listing all users and channels.
- Adding Slack adds 8 tools, Notion ~20 — every additional tool degrades reasoning; pagination loops further bloat the context window.
- Solution: blur the line between tool and agent — give tools more agency so they accept natural-language inputs, do their own reasoning, and return clean outputs (Avengers metaphor vs T-Rex-with-spanner).
- Demos Wordware's MCP Toolbox: package agentic flows (e.g. competitor analysis hitting Twitter scrape + Notion) as a single high-level tool for another agent.
mcptool-designagents
Original description
There's a new type of wrapper in town. The MCP API wrapper. Make them thin and you'll be wondering why your chatbot is struggling to even send a Slack message (true story). But make them _agentic_ and the world is unlocked. In this talk I'll demonstrate the drawbacks using low level APIs as MCPs and show the magic that happens when your 'tools' are actually other agents. It's prompts all the way down baby!