One agent.
Every clock.
Every stack.
The TimeBeat software platform is built on open standards, deployable on any Linux or Windows host, observable from the tools you already use, and designed so your ops team can run it in production on day one — not week six.
1
Agent binary
Open
Source stack
Self-host
or cloud
timebeat-install.sh
# install the agent
$ curl -sSL install.timebeat.app | sh
# point at your backend
$ timebeat configure \
--backend timebeat.corp.local
# go
$ systemctl start timebeat
✓ agent running, reporting to backend
Architecture
Four layers. One mental model.
A lightweight agent on every host. A scalable backend that stores and indexes. A dashboard for humans. APIs for everything else. That’s the whole system — and nothing in it is opinionated about the rest of your stack.
01
Agent
One binary per host — Linux or Windows. Auto-discovers PTP, NTP, GNSS, PHC devices. Streams metrics over gRPC (mTLS).
02
Backend
Time-series store tuned for clock metrics. Retention, indexing and compaction handled. Horizontally scalable.
03
Dashboard
Topology, live metrics, historical replay. Every clock, every PTP domain, every grandmaster in one view.
04
APIs
REST and gRPC for every metric, alert and configuration. OpenAPI spec ships with every release.
Jitter attenuation
The signal processing that cleans your clock
Every PTP and NTP timestamp the agent captures carries noise — network jitter, asymmetric queueing delay, oscillator phase noise, GNSS multipath and interrupt latency on the host. Raw timestamps are unreliable. The TimeBeat jitter attenuation engine processes them into a clean, stable clock signal that your downstream applications can actually trust.
This is the core differentiator of the TimeBeat software stack. Other tools display raw phase offset. TimeBeat attenuates the jitter first, then displays the result — which means the clock your operations team sees is the clock your applications experience, not the noisy raw signal that nobody can act on.
Multi-source filtering
Aggregates timestamps from multiple PTP masters, NTP servers and GNSS receivers simultaneously. Cross-correlates the sources to identify and suppress outliers — a single degraded source doesn't contaminate the output.
Adaptive loop bandwidth
Dynamically adjusts the servo loop bandwidth based on observed noise characteristics. Tight bandwidth during stable GNSS lock for maximum filtering. Wider bandwidth during holdover entry so the oscillator tracks gracefully without overshoot.
Statistical outlier rejection
Every incoming timestamp is scored against a running statistical model of the expected offset and delay. Packets with anomalous delay — caused by transient network congestion, queueing spikes or asymmetric path changes — are rejected before they reach the servo.
Sub-nanosecond phase noise floor
After attenuation, the residual phase noise on the disciplined clock is below one nanosecond RMS in steady state on hardware-timestamped deployments. This is the precision your downstream application sees — not the raw PTP offset, but the attenuated output.
Holdover-aware transition
When the primary reference degrades or is lost, the attenuation engine smoothly transitions to the local oscillator's free-run characteristics rather than abruptly switching. The result is a holdover entry that the downstream application can't distinguish from normal operation.
< 1 ns
Residual phase noise (RMS, steady state)
100×
Typical jitter reduction vs raw PTP
< 50 ms
Holdover transition (indistinguishable to applications)
Multi-source
Simultaneous PTP + NTP + GNSS filtering
Supported platforms
Linux or Windows — it runs TimeBeat
Major Linux distributions, Windows Server and Windows desktop, container runtimes and Kubernetes. We don’t ask you to change your platform to run ours.
Ubuntu
20.04 LTS +
Debian
11 +
RHEL / Rocky / Alma
8 +
SUSE
15 +
Windows Server
2019 +
Windows
10 / 11
Docker / OCI
Any runtime
Kubernetes
1.24 +
Bare metal
kernel 5.4 + (Linux) or Windows Server 2019 +
systemd / OpenRC / Windows Service
Any
Integrations
Plugs into the stack you already run
TimeBeat isn’t a replacement for your observability stack — it’s a clock-specific layer that streams into whatever you already use for metrics, alerts, dashboards and logs.
OpenTelemetry
Traces + metrics
Grafana
Dashboards
Datadog
Observability
Splunk
Logs + SIEM
PagerDuty
Alerting
Slack
Notifications
Kafka
Event streaming
S3-compatible
Cold storage
Webhook
Custom integrations
Open by design
Standards-based and interoperable
The TimeBeat Agent is our own synchronisation engine — not a wrapper around someone else’s daemon. It implements every major PTP profile and interoperates with any standards-compliant network, so it drops into your fabric with no lock-in. Sync Insight then gives you complete observability over the agent — every clock, every offset, every event — exposed through open APIs that fit the tooling you already run.
APIs
Every metric, every alert, every config — via API
REST for easy integration. gRPC for performance-sensitive paths. OpenAPI and Protocol Buffers schemas ship with every release.
curl — REST
$ curl -H 'Authorization: Bearer $TOKEN' \
https://timebeat.corp.local/api/v1/clocks/primary
{"id":"primary","offset_ns":-43,"gnss":{"quality":"locked",
"sats":18},"holdover":false,"source":"gnss"}
grpcurl — gRPC
$ grpcurl -H 'authorization: Bearer $TOKEN' \
timebeat.corp.local:9443 \
timebeat.v1.ClockService/StreamMetrics
{"clock_id":"primary","offset_ns":-43,"ts":"2026-04-11T12:00:00Z"}
{"clock_id":"primary","offset_ns":-38,"ts":"2026-04-11T12:00:01Z"}
Deployment
Three ways to run it
Whatever your security model, compliance posture, or ops appetite — there’s a way to deploy TimeBeat that fits.
Self-hosted
Your infrastructure. Your data.
Deploy the backend on any Linux or Windows host, or on Kubernetes. Your telemetry never leaves your network. Full RBAC, SSO (SAML / OIDC) and audit logging. The default choice for banks, defence, and regulated industries.
- Single-binary backend
- SAML / OIDC SSO
- S3-compatible storage
- Air-gap friendly
TimeBeat Cloud
Managed. Multi-region. Zero ops.
The full platform, hosted. Global PoPs, 99.999% SLA, continuous updates, no backend infrastructure to run. Ideal for teams that want the capabilities without the ops overhead.
- 99.999% uptime SLA
- Global PoPs
- Continuous updates
- Zero backend ops
Air-gapped
For classified and isolated networks
Fully offline installation for networks that can’t touch the public internet. Offline license activation, offline updates via signed artifacts, full platform capability with zero external dependencies.
- Fully offline install
- Signed update bundles
- No external telemetry
- Classified-network tested
Security & trust
Built for regulated environments
Banks, defence and critical infrastructure demand more than "it encrypts in transit". So we default to everything they need.
Encryption in transit and at rest
All agent → backend traffic is mTLS-encrypted. All data at rest is AES-256 encrypted with customer-managed keys where required.
RBAC + SSO
Fine-grained role-based access control. SAML, OIDC and SCIM integration with Okta, Azure AD, Google Workspace and your IdP of choice.
Audit trail
Every configuration change, every alert acknowledgement, every API call logged to an append-only audit stream. SOC 2-ready.
Signed releases
Every binary is cryptographically signed. Every container image is in a verified registry. Tamper-evident all the way to the kernel.
And what does it actually do?
The six capabilities of the platform
Observability, analytics, UTC verification, resilient topology, virtual grandmasters, and multi-source clock fusion — all built on the architecture above.
See it live
Request an engineering demo
Thirty minutes with a TimeBeat engineer. We’ll walk you through a live environment, answer your architecture questions, and map the platform to your existing stack.
“TimeBeat gave us the timing observability we’d been missing. We caught three drift events in the first week that our legacy monitoring never would have.”
— Network Architect, Tier 1 European Bank
Built by engineers, for engineers