Open-source toolkit for agent leads
AgentLead is two open-source tools for the human in the loop. Skills make the agent in your editor follow real engineering discipline. Flow lets you stand up and direct your own team of agents — declared in YAML, with durable state and clear lines of authority.
Two ways to lead
One standard of rigor, whether the agent is writing your code or running in production.
Your AI coding assistant, led like a disciplined engineer. You don’t pick a skill — you describe the task and the right flow fires: investigate, design, test-first, review, verify.
Stand up your own agents in YAML — each a system prompt plus tools — and run them as a graph over shared state. You hold the authority; the model only gets what you grant it.
The throughline
The judgment that makes a feature ship cleanly — investigate before you build, design the interface first, test what matters, verify before you merge — is exactly what a team of agents needs before you can trust it with real work.
Skills instills it in the assistant at your side. Flow lets you encode it into the agents you deploy — explicit control flow, durable state, and a hard boundary on how much authority the model holds.
# direct the agent in your editor
$ npx @ai-agent-lead/skills
# command your own team of agents
$ pip install agentlead-flow-core
$ agentic-flow pipeline.yaml --tools my.tools
# same standard, every agent on the team.Both are open source, free, and a single command away.