Stop AI agents before they drop your production database.
Block the catastrophic deterministically.Coach the recoverable.
It physically can't run the destructive call—even when your agent tries. With your own LLM key, the block becomes a coached recovery: the agent revises and finishes instead of dying on a 403.
Start free, on your machine—no LLM key, nothing leaves your laptop. Add your own LLM key to unlock Recover; connect the cloud control plane for team HITL & SOC.
Live across the protected network
--
Irreversible / exfiltration actions stopped pre-execution.
--
Share of challenged agents that self-corrected.
--
Runs that self-corrected vs. crashed.
--
~est. 20 min saved per protected run.
pip install agentx-security-sdkfrom agentx_sdk import agentx_protect, is_block
@agentx_protect(agent_id="crm_worker")
def dispatch_update(client_id, notes, db=None):
# never reached — the shield blocks the call first
return db.execute(notes)
# a DROP TABLE smuggled into the notes via prompt injection:
out = dispatch_update("c-99401", "...; DROP TABLE users;")
is_block(out) # -> True · blocked locally · no keyThat's the keyless Layer-0 shield—it stops the blatant catastrophic calls (DROP TABLE, secret-exfil, SSRF) right in process RAM, no key required. Only dependency-reputation checks ever reach a public registry.
Next → drop in your own Gemini key and the block becomes a coached recovery: the agent revises and finishes the task instead of dying on a 403. That's Recover.
Ready for the full deterministic floor (AST parsing, SSRF normalizer, the whole failure catalog) plus team HITL/SOC? Run the gateway:
docker compose up -d # the full floor runs hereThe gateway image ships with design-partner access—request it below.
pip install+ run the gateway. The full deterministic floor hard-blocks your agent's catastrophic call before it runs—no LLM key, no security team, runs on your machine.
block + nudge
The block becomes a coached, recoverable challenge—your agent revises and finishes the task instead of dying on a 403. Keep shipping, unattended.
guide + continue
Connect the cloud control plane for team human-in-the-loop & SOC approvals, shared dashboards, and a fleet-wide audit trail—central oversight for when one machine isn't the whole story.
review + govern
Request access—we'll send your gateway keys to run the floor locally, plus the cloud control plane (team HITL, SOC & shared dashboards) when you're ready.