ResearchAgentman originalsAutonomous agents
Agentman Alpha
by Agentman
Our flagship bundle, and the one that shows what changes when a plugin ships autonomous agents instead of skills alone.
What’s in the bundle
- Autonomous research agents that pursue a question across sources
- Orchestration agents that sequence multi-step work
- Action agents that operate against connected tools
- The skills the agents call, usable directly when you want the control
What you can do with it
- Hand over a research question and get back a sourced answer, not a search
- Run a multi-step workflow that would otherwise need supervision at each step
- Chain work across tools without wiring the integration yourself
What it doesn’t do
Autonomy is the feature and the risk. Agents that act need scoped permissions and a review step on anything irreversible — start them on read-only work and widen the scope once you have watched how they behave.
Where it runs
Agentman Alpha ships as MCP-compatible skills, so it works in Claude, ChatGPT, Cursor, Claude Code, and Agentman over a single connection — no per-tool setup.
Browse all Claude plugins