Controlled beta · Alert-only by default

Your servers don’t need another dashboard.

They need a safer first responder. AgentPulse detects repeat host incidents, prepares a bounded fix, checks local policy, and proves whether the action worked.

Request a controlled pilot See the decision loop

Local authorityThe host decides what is allowed.

Verified actionsEvery change is checked afterward.

No remote shellKnown actions, not arbitrary commands.

LIVE DECISION TRACE LOCAL POLICY ACTIVE

INCIDENT / DISK PRESSURECache volume at 91%

00:14:32

ObserveThreshold crossed · 91.2%

DONE

Simulate1.8 GB recoverable · 0 protected

SAFE

GatePath allowlist + symlink boundary

PASS

ActAwaiting approved mode

ASK

VerifyRe-check original condition

NEXT

BUILT FOR SMALL INFRASTRUCTURE TEAMS

Linux

macOS

systemd

launchd

Generic webhooks

01 / THE PROBLEM

Monitoring sees the fire.
AgentPulse handles the known first move.

A full observability stack can show every spike, trace, and log line. It still leaves someone SSHing into a host to restart the same service or clear the same safe cache path.

AgentPulse is deliberately narrower: repeat host incidents, explicit authority, deterministic safety gates, and verification after every attempted fix.

02 / THE SYSTEM

Authority stays local.
Evidence travels.

Cloud policy can make the agent more restrictive. It cannot widen the authority configured on the host.

Explore the safety model →

HOSTLocal agentChecks · policy · actions

BOUNDARYLocal allowlistCannot be widened remotely

CLOUDFleet evidenceStatus · history · escalation

OUTBOUND-ONLYREDACTEDRETRY-BOUNDED

03 / CONTROL

Not “autonomous” or “manual.”
Exactly as much authority as each check deserves.

Promote one low-risk action without giving every incident the same permission.

01Off

Do not run this check.

02Alert

Detect and notify. Change nothing.

DEFAULT

03Ask

Prepare the fix. Wait for approval.

REVIEW

04Auto

Run an allowlisted fix, then verify.

OPT-IN

04 / REPEAT INCIDENTS

Built for the incidents you’re tired of fixing twice.

Known problem. Narrow response. Measurable outcome.

Disk pressure

Clean only configured paths. Refuse system paths, directories, and symlink escapes.

Clean → re-check

Crashed service

Restart only an allowlisted systemd or launchd service, then confirm it is active.

Restart → verify

Runaway process

Identify the largest memory offender and preserve the termination decision for a person.

Detect → report

Fix did not hold

Record failed verification and escalate instead of entering a destructive retry loop.

Stop → escalate

PRODUCT BOUNDARY

The confidence comes from what AgentPulse refuses to do.

No arbitrary browser-to-host command channel.

No automatic process killing in the current release.

No widening local authority from the cloud.

No blind retries after verification fails.

Review the current release boundary →

05 / FOUNDING BETA

Start with one host.
Earn the next permission.

Every pilot begins on one approved non-critical host in alert-only mode. Multi-host plans remain founding reservations until the full paid lifecycle is ready.

Request beta access↗

STARTERControlled pilot

C$29/ month CAD

One host · alert-only first

PRO · FOUNDINGSmall fleet

C$99/ month CAD

Up to 5 hosts when fleet access ships

BUSINESS · FOUNDINGAgreed fleet

C$299/ month CAD

Finite host limit confirmed before billing

Reservations are free. No charge until the matching service is ready and you choose to activate.

ONE REPEAT INCIDENT IS ENOUGH TO START

Stop being the remediation layer.

Tell us the host type, stack, and incident you keep fixing. No credentials. No server addresses. Just enough to confirm fit.

Request a controlled pilot↗ [email protected]