Collaborative Redteam [Preview]
Collaborative Redteam is the interactive, multi-user surface of Red Coworker. Where automated campaigns produce reports unattended, Collaborative Redteam puts humans back in the loop — your security team and the AI work together on live attack sessions against your agent.
Purpose
- Explore specific attack angles that automated tests don't cover
- Pivot in real time based on how your agent responds
- Train your team on adversarial techniques against your specific agent
- Share session state across teammates so multiple people can collaborate on the same engagement
How it differs from automated tests
| Automated tests | Collaborative Redteam | |
|---|---|---|
| Mode | Fully automated | Interactive, human-directed |
| Duration | 20-90 minutes (unattended) | As long as you want |
| Strategy | Pre-defined OWASP categories | You choose + AI suggests |
| Adaptability | Automated backtracking | You decide when to pivot |
| Output | Full report | Per-session verdicts + findings |
Findings discovered during a session flow into the same lifecycle as automated findings — they get assigned, tracked, and verified through the Red Coworker collaboration loop.
Preview
Collaborative Redteam is under active development — moving from CLI to a webapp interface. The interaction model, interface, and integration points may change as we refine the experience based on feedback.