Start free. Upgrade when you ship at ACT-3/4.
Free tells you something may be wrong. Pro tells you whether you can ship.
…under what controls, and with whose accountability.
"Is this agent safe to merge?"
Pro reviews the implementation in-editor and tells you exactly which controls are missing before the PR, not after the incident.
"Can I prove this maps to our frameworks?"
All 32 frameworks, full AAF scoring, and CP.9/CP.10 give you the evidence a regulator or auditor will ask for.
"Who is accountable if it goes wrong?"
ACT-3/4 governance with the HEAR doctrine names human authority at the execution boundary — before consequential actions complete.
Common questions, straight answers.
Which tools does it work in?
+
Claude Code, Cursor, and Codex — the MCP runs inside the AI development tools your engineers already use. No separate console, no context switching.
What does the free tier include?
+
A free token gets you agent classification and baseline governance checks inside your tools. Pro unlocks the workflows you need when you ship agents at ACT-3/4 autonomy — the full comparison table is above.
What is an ACT level?
+
AI SAFE²’s agent classification tiers. Higher tiers mean more autonomy — and more governance required before production. The MCP classifies your agents and tells you which tier you are actually operating at.
How do we install it?
+
One command: npx @aisafe2/mcp install. Request a free token and setup instructions arrive by email.
Simple tiers. Pro is the rational middle.
All 32 frameworks, AAF scoring, code review, ACT-1–4, CP.9/CP.10, hosted endpoint.
Buy Pro