A single binary that takes active control of the host system clock, replacing chrony / ntpd / ptp4l / the Windows Time Service. Every major PTP profile, five steering algorithms, Clock Ensemble fusion, PTP² Mesh topology, and 167 fields of telemetry out the box. Linux and Windows. Physical, virtual or cloud.
Looking for the platform overview? → Sync Insight Platform
What the Agent is
The Timebeat Agent is the active synchronisation engine on the host. It reads from your configured time sources, runs a steering algorithm against the local clock, writes adjustments to the kernel, and emits telemetry. It is explicitly the sole synchronisation daemon — chrony, ntpd, ptp4l, the Windows Time Service all have to be disabled before the Agent starts. Only one agent controls a system clock.
Live sync cycle · algorithm, offset, source weights, oscillator state
Agent feature tour
Each feature summarised here has an extended engineering guide in our Resources library. The summaries are the menu; the guides are where the protocol details, configuration examples and operational considerations live.
Active synchronisation
Five selectable algorithms (sigma, rho, alpha, beta, gamma) steer the host clock against the chosen reference. Outlier filter with strict/moderate/relaxed modes and per-peer opt-outs. Offset-arrival-driven servo mode for bursty sources. Interference monitor detects competing time agents and backs off rather than fighting them. Full servo PID state (P, I, D, total adjustment, steering type) logged every cycle.
Steering & servo telemetry →Multi-source fusion
Fuses GNSS, upstream PTP feeds, PPS inputs and local oscillator into a single weighted output — statistically tighter than any single input. BIPM-style weighted averaging with automatic outlier rejection. A degraded source is de-weighted smoothly rather than triggering a failover event. The fused output is what downstream clients see.
Clock Ensemble — deep dive →Self-healing topology
A peer-to-peer overlay that runs between Timebeat Agents and coordinates who serves time to which downstream clients. mDNS on the LAN, DHT across segmented networks. Seat-based capacity model, active-active load balancing, sub-second failover when an Agent degrades. Downstream clients see one stable virtual grandmaster, not the backing pool.
PTP² Mesh — deep dive →Multi-tenant time feed
One Agent serves many independent virtual grandmasters — a dedicated, securely segregated PTP time feed for each tenant out of shared hardware. Every tenant gets its own isolated grandmaster and addressing, so tenants with overlapping IP ranges never collide and you don't burn a physical switch port or a dedicated grandmaster box per tenant. Backed by a Mesh-coordinated pool with sub-second failover, it's the default pattern for multi-tenant Shelf deployments, secure segregated networks and managed service provider platforms.
VGMC pattern — deep dive →Precision timestamping
SOF_TIMESTAMPING hardware kernel-bypass on supported NICs — essential for sub-100 ns accuracy. Per-interface PHC offset strategies (precise, efx/sfc, extended, PPS, basic); modified lucky-packet and non-Gaussian PHC outlier filters; 1-step PTP mode where NICs support it. External software timestamping fallback for environments without hardware support.
Hardware timestamping telemetry →Protocol support
IEEE 1588-2008 and 2019. G.8275.1 / G.8275.2 telecom, G.8265.1 legacy, enterprise-draft, IEC/IEEE 61850-9-3 power utility, SMPTE ST 2059-2 broadcast, AES67 audio. NTP v4 with configurable poll intervals. 1 PPS hardware inputs. NMEA serial GNSS. PHC inputs including Azure Hyper-V and VMware virtual PHCs. White Rabbit fibre. Fallback and oscillator-discipline source protocols.
PTP profile selection guide →Compliance verification
Every second, the Agent verifies the host clock — cross-checking around 167 telemetry metrics across nine measurement domains plus the Agent's own configuration to prove the clock is provably in spec, not merely steered toward it. The result is a continuous, auditable record of UTC traceability that satisfies DORA Article 9, MiFID II RTS 25 traceability and SEC CAT evidence requirements simultaneously. Available at the Sync Insight Enterprise tier.
UTC Verification — white paper →Observability output
Every synchronisation cycle, the Agent emits around 167 telemetry fields across nine measurement domains — clock offset, servo state, PTP topology, upstream source health, hardware timestamping state, GNSS signal quality, OSNMA authentication, position integrity, hardware sensors. ECS-compliant structured events; native Elasticsearch, OpenTelemetry, REST and gRPC.
167 telemetry fields — engineering guide →What flows out
Every cycle the Agent emits structured telemetry covering the full state of the sync stack. Most fields are diagnostic — you never look at them in normal operations — but when something goes wrong the relevant field is the difference between a two-hour incident and a twenty-minute incident.
167 fields · ECS events · 1 Hz · Elasticsearch / OpenTelemetry
Deployment model
The Agent runs on Linux and Windows, on physical servers and in virtualised or cloud environments. It integrates with systemd on Linux and Windows Services on Windows. No dependency on Sync Insight or Timing Feed — those are higher-level products; the Agent is the foundation that can run on its own.
Linux deployment
/etc/timebeat/timebeat.yml./var/log/timebeat with rotation + structured ECS output.Windows deployment
C:\ProgramData\Timebeat\timebeat.yml.YAML config — primary / secondary clocks, PTP profile, Ensemble, output
Licensing
With Sync Insight
For every Sync Insight subscriber, the Timebeat Agent is included on every monitored host with no separate licence and no per-host fee beyond the Sync Insight pricing. All premium Agent features (Clock Ensemble, PTP² Mesh, UTC Verification) are activated by the tier.
Sync Insight pricing lanes →Standalone
The base synchronisation engine is free — download and run. Premium features (Clock Ensemble, PTP² Mesh, TaaS / VGMC, UTC Verification) are gated behind a licence key that points at your /etc/timebeat/timebeat.lic. For standalone enterprise deployments, contact sales for a quote.
Library
Extended deep dives on the features summarised above. Each guide is the source of truth.
How PTP² Mesh turns a fleet of Timebeat Agents into a self-discovering, self-healing time distribution network. mDNS and DHT peer discovery, seat-based capacity, active-active operation and millisecond failover — for when you need redundancy without the rigidity of classical BMCA hierarchies.
A virtual grandmaster clock is an IP endpoint that looks like a single PTP grandmaster to downstream clients but is backed by multiple physical Timebeat Agents — redundancy, capacity and failover at the topology level, with a single client-facing configuration.
What jitter attenuation actually does inside a PTP timing fabric, why raw timestamps are unreliable, and how TimeBeat's signal-processing engine delivers sub-nanosecond residual noise from noisy real-world inputs. Written by TimeBeat's engineering team.
Why a single-grandmaster deployment is a DORA Article 11 problem, what A/B/C redundancy looks like in a single rack unit, and how the Open Time Appliance Shelf turns three independent Rubidium Black+ grandmasters — with independent GNSS antennas — into the default finance-venue topology for 2026 and beyond.
A speculative question, not a product pitch: imagine if every machine could hand anyone independently verifiable proof of what its clock was doing at any moment in history. Would it matter? Who would benefit, what might it cost us, and would the world actually be better for it? An exploration of the idea of cryptographic UTC attestation.
An engineering-level tour of the 167 telemetry fields the Timebeat Agent emits per cycle to Sync Insight. Nine measurement domains, why each one matters for operations or compliance, and how to pick the handful of fields your Grafana dashboard actually needs day-to-day.
Deploy the Agent
The Agent installs in minutes on a single host, replacing your existing synchronisation daemon. Run it standalone for a week to evaluate; add Sync Insight for observability when ready.
Every host · one daemon · eight features