The free scanner is live. See your agent's blast radius in one command.
Hermes Shield is the actions firewall for AI agents. The free scanner maps your agent's action surface locally and read-only, so you see your blast radius before you trust anything. The paid Repairer and Shield tiers are coming.
pip install hermes-shield-scannerthen hermes-shield demo · v0.8.2 · live on PyPI · GitHub
then run hermes-shield scan in your own repo
Independently security-reviewed by an industry professional.
Find it free. Fix it. Shield it.
One suite, one honest ladder — Discovery → Repairer → Shield → Enterprise. The Discovery scanner is live and free to install today. The Repairer and Shield tiers are coming; teams and enterprise can register now.
- Local, read-only scan of your repo
- Full action-surface report
- Public-scans access
- Everything in Discovery
- One-time hardening of your findings
- Action Gates + Threat Shield config
- Everything in Repairer
- Live Kill Switch + Action Gates + Audit Trail
- Drift monitoring + ongoing support
- Everything in Shield
- Local-first deployment path
- Security-team walkthrough + evidence packs
▸ Free scanner, live now · install and run hermes-shield scan · local, read-only by default · Repairer and Shield coming.
You pay for control, not for alarms.
We audited 12 agent frameworks — a combined 361k+ GitHub stars across the frameworks we scan — and mapped 8,447 action surfaces. Of those, 542 are install-liability and 596 are reachable-in-repo — two distinct tiers, not a subset of each other. That inherited action surface is the risk OWASP names LLM06 · Excessive Agency. Detection tells you about it afterwards. The coming Shield tier is built to stand in the action path: Map → Trace → Prove → Block → Patch, on your box, on your terms.
Discovery · Free · Live now
- Local, read-only scan of your repo
- Full action-surface report
- Public-scans access
Repairer · Coming
- Everything in Discovery
- One-time hardening of findings
- Action Gates + Threat Shield config
Shield · Coming
- Everything in Repairer
- Live Kill Switch, Gates & Audit Trail
- Drift monitoring + ongoing support
- Enterprise: coming, per team
We price it the way we bought it.
Our own agent got prompt-injected. We built the Shield for ourselves first, ran it free, and only started paying attention to packaging once it was blocking real actions. Your ladder mirrors ours.
What we will not do to your bill.
No egress tax
A default scan runs fully local and reads your code read-only — nothing is sent anywhere, so there is no data-egress line item. The optional --ai tier uses your own local Claude CLI; that is the only mode that sends any code off your machine, and only to Anthropic's API under your own Claude account.
No per-alarm pricing
Detection tools charge by the noise they make. The coming Shield tier controls actions — the unit of value will be what gets blocked, traced and patched, not the noise.
No lock-in theatre
When the Shield ships, its Audit Trail, Action Gates and Kill Switch will produce evidence you own. Leave whenever you like; the evidence stays with you.
Scan first. Decide with evidence.
Install the free scanner, run hermes-shield scan on your own agent's codebase, and see exactly what the Shield would be standing in front of, locally and read-only, before a single pound changes hands.
pip install hermes-shield-scanner▸ Want the paid Repairer or Shield tier? Join the waitlist. Independently security-reviewed by an industry professional.