The download's checksum doesn't match#
Don't run the file. Delete it and download it again with the command shown in Enroll agent; a mismatch usually means an interrupted or altered download. If it still doesn't match, contact [email protected].
Documentation
Fixes for enrollment, offline agents, refused commands and scans, plan limits and sign-in.
Don't run the file. Delete it and download it again with the command shown in Enroll agent; a mismatch usually means an interrupted or altered download. If it still doesn't match, contact [email protected].
The enroll panel couldn't get the current release information, or there is no build yet for the platform you picked. Try again in a few minutes or pick another platform. If it persists, contact support.
Each enrollment token is valid for 20 minutes and one machine, and is shown only once. Generate a new one in Enroll agent and run the commands again with it.
Run the enroll and install commands with sudo on Linux and macOS, or in a PowerShell window opened with "Run as administrator" on Windows. Enroll must succeed before install: run the commands in the order shown.
The agent needs osquery to read system state, and install checks for it first so the service doesn't start into a restart loop. Install osquery from osquery.io, then run the install command again.
The machine already belongs to an organization. To enroll it again (for example into a different organization), first run the agent's uninstall command, then unenroll, both as root or administrator, then enroll with a new token.
Check that the agent service is running on the host. On Linux: sudo systemctl status pcs-agent (and sudo journalctl -u pcs-agent -n 50 for its recent log). On Windows: Get-Service pcs-agent in PowerShell.
If the service is running, make sure the host can reach the Sokendi server. After a network outage the agent reconnects on its own, waiting up to about 30 seconds between attempts. An agent that was revoked can't connect again; enroll it with a new token instead.
A command runs only once the agent is online. If it's offline, the command waits and is delivered when the agent reconnects, unless its timeout passes first, in which case it expires. Check the agent's status on the Agents page; the command's status shows under Last result on the Issue command tab, and in the agent's History tab.
Remote actions are off until a Tenant Admin turns them on: Agents → Remote actions → Allow remote actions, then tick the action. Ad-hoc osquery only works for tables on that list, and remediation only for the types ticked there. After saving, agents apply the change on their next policy check, usually within a minute or two; if a command still fails right after a change, wait a moment and issue it again.
Every table a query reads must be on your organization's list, including tables in joins, subqueries and "IN table". The message names the tables that aren't. A Tenant Admin can add them under Agents → Remote actions → Ad-hoc osquery → Edit.
A few things are always refused whatever the list says: curl, curl_certificate and carves (they make the host send network requests or upload files), names with a schema prefix such as main.processes, more than one statement, and anything other than a SELECT.
The server can't check queries at the moment, so it refuses all of them rather than letting an unchecked query through. Rescan and remediation still work. Try again later, and contact [email protected] if it continues.
The agent always refuses PID 0 and 1, its own process and osquery helper, and processes whose termination would crash Windows (such as lsass.exe or csrss.exe). It also refuses a PID that no longer exists: process IDs change when programs restart, so run the query again and use the PID from the new result.
The agent checks every command before running it: that it is signed by the platform, addressed to that agent and not expired. A rejection usually means the command expired before the agent received it (issue it again), or the agent's credentials no longer match (re-enroll the agent). Repeated rejections you didn't cause are worth reporting to support.
Hash indicators reach agents through their normal policy fetch, which can take a few minutes. Check that the hash is a SHA-256 (exactly 64 hexadecimal characters) and that the agent is online. Matches are flagged on running processes, so the file must actually be running on the host.
Detections, like scan results and agent telemetry, are deleted automatically once they are older than your plan's retention period (30 days on Free, 45 on Standard, 90 on Pro, 365 on Enterprise), whatever their status.
The map only plots IP addresses that are public and routable (private LAN addresses on your own agents are left out on purpose) and that have appeared in network activity promoted to the live feed. An empty map alongside real threat feed entries usually means none of them involve a public remote address yet.
This is expected. Enrolled agents almost always report a private LAN or VPN address (for example 192.168.x.x), which has no real-world location and is never placed on the map. Only the Attackers filter (public remote IPs seen in agent network activity) shows points for agent data. "My Agents" is for the rare agent that has a public IP itself.
If routine activity shows up as Critical, the detection settings are probably too sensitive for your environment rather than something being wrong. A Tenant Admin can review the detection settings under Threat Intel; contact support if it continues after a recent agent update.
Active scans only run against domains your organization has verified by publishing the DNS TXT record Sokendi gives you.
Verification is re-checked every week. If the record is removed, or can't be found for 14 days after repeated checks, the domain goes back to unverified and scans against it are refused until you verify it again. Keep the _sokendi-challenge TXT record in place.
On Pro and higher, active scans need business verification. Open Settings → Verification: it lists what's still missing. Usually that's the business details (a Tenant Admin adds the legal business name) or an admin email on a verified domain: at least one Tenant Admin must sign in with an address on one of your verified domains, so if all your admins use personal addresses (such as gmail.com), invite one with a company address. Passive lookups (WHOIS, IP geolocation, CVE lookups) don't need it.
Each plan can verify a set number of domains: 3 on Standard and 15 on Pro. Every domain you've added counts, pending or verified, so remove one you no longer need under Verified domains, or upgrade. On the Free plan, adding domains isn't available: Free runs passive lookups only.
This happens after moving to a plan that allows fewer domains. Your domains stay listed, but only as many as the plan allows can be scanned, starting with the ones verified first. Remove the domains you no longer need, or upgrade, to scan the others again.
The Free plan includes 25 scans per day. Standard and Pro include unlimited scans under fair use, with a daily safeguard against automated or runaway volume that normal use doesn't reach. Both reset at 00:00 UTC. A scan counts when it starts or is retried, and running a single module from the Modules page counts too. If you reach the fair-use limit during normal work, contact [email protected].
Each plan can run a set number of scans at the same time: 1 on Free, 2 on Standard and 5 on Pro. Wait for one to finish, or abort one you don't need, then start the next.
Breach checks only cover email addresses on your organization's verified domains, so an address on any other domain (including personal ones such as gmail.com) is refused. Verify the domain first, or check an address on a domain you've already verified.
Active scans run against verified domains only, because an IP address can't be proven owned with a DNS record. Scan the domain that points to the address instead. Scanning IP ranges can be arranged on Enterprise under a signed scan authorization; contact [email protected].
After repeated failed sign-in attempts, sign-in for that account is paused for 15 minutes. Wait, then try again, or use Forgot password. Changing your password signs you out on every device.
Managing team members requires Tenant Admin, and the last remaining Tenant Admin can't be demoted or removed, so an organization is never left with nobody able to manage it. Promote another member to Tenant Admin first if you need to change that.
Check the Team page first: if their email shows as "Not verified", the invite went through but they haven't finished setting up. Ask them to check spam or junk, and confirm the address was entered correctly; inviting a corrected address counts as a new invite, not an edit.
Only a Tenant Admin can change the plan. If you're a Viewer or Analyst, ask your organization's admin to make the change, or to give you the Tenant Admin role if you're meant to manage billing. Platform administrators don't see this option either, since it belongs to an individual organization's subscription.
This is expected: only Free → Standard → Pro upgrades are self-service. Enterprise plans (custom limits, dedicated support) are set up by the Sokendi team. Email [email protected] to discuss one.
Deleting an organization deactivates it rather than cutting every connection at once, so agents may keep trying to report for a short while before they are refused. Uninstall the agent from those machines. The organization's data is kept for 30 days so it can be restored if the deletion was a mistake, then permanently deleted; contact support within that time if you need it back.
Email [email protected]. Response times by plan are on the Support page.