Sysdig AI Defense
Sysdig's runtime control for AI agents that checks each action against policy and can allow, hold for approval, or block it.
Overview
Sysdig announced general availability of Sysdig AI Defense on November 5, 2026. It builds a live inventory of agents, harnesses, models, MCP servers, skills and plugins from runtime telemetry, naming Claude Code, Codex, GitHub Copilot and Cursor, and ties each agent to an accountable person. Each agent action is evaluated before it completes, and sessions are recorded with a live action graph. It is built on kernel-level instrumentation and the open source Falco project.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- Live inventory of agents, harnesses, models, MCP servers, skills and plugins
- Pre-execution policy: allow, hold for human approval, or block; templates mapped to OWASP Top 10 for LLM Applications
- Policies start in alert mode and move to enforcement
- Session recording with an action graph comparing stated intent to actual behavior
- Immutable audit records of which policies applied and who approved
- AI spend attribution by agent, team and person, including calls that bypass an AI gateway
Use cases
- Blocking unvetted MCP servers, skills and plugins on developer workstations
- Holding risky credential reads by an agent for human review
- Auditing what a coding agent actually did during an incident
Pricing
Unknown; not stated in launch coverage.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 5 Nov 2026
- Record updated
- 7 Nov 2026
- Record last checked
- 7 Nov 2026