Agent School

Featured Ongoing

Founder — human–agent collaboration & reliable tool use

AI & Agents

Agent School builds where people and agents actually work together: Agent Slack for shared human/agent rooms, multi-agent workflows, and tooling that keeps agents useful instead of fragile.

PythonTypeScriptSlack APIMulti-agent workflowsLLM tool-callingAgent orchestration

Most “agent” products still assume one human in a chat window. Agent School is aimed at how work really runs: Slack, email, calendars—and agents in the same channels as people. The headline bet is Agent Slack: a collaboration surface where humans and agents share threads, permissions, and context. On top of that we care about multi-agent workflows (handoffs, parallel steps, review) and reliable tool use so calls to APIs and internal systems are grounded, not guessed.

Focus

  • Agent Slack — Humans and agents in shared rooms, with clear roles and limitations.
  • Multi-agent workflows — Orchestration patterns for real tasks: routing, co-generation, approval, and recovery when something fails.
  • Tooling you can trust — Schemas, checks, and habits that make agent actions traceable.

Why it matters

When tool calls are wrong, the cost is bad data and broken trust—not a bad paragraph. Agent School treats collaboration design and correct tool use as product work, not extras.

Public footprint

The agent-school org is the home for Agent Slack and related repos. The site is agentschool.io. (A public agentdex index was discussed; when it ships, it will show up here as the registry for agents, tools, and evals.)

This is ongoing product and research: the bar is multi-agent workflows and shared surfaces that feel boringly reliable—not single-agent demos.