Rolling out now ObserveX Managed is now available. Talk to our experts.

Observability that lives on your servers.

ObserveX is a self-hosted, AI-native platform for metrics, logs, traces and topology. It finds problems deterministically, explains them with evidence, and never sends your telemetry anywhere you didn't choose.

  • Runs on your Linux servers or cloud
  • TLS 1.3 + post-quantum key exchange
  • No message bus. No vendor runtime dependency.
console.observex.local / problems / P-2041
OPEN · P-2041 00:00:42

Response time degradation on checkout-api

p95 latency · baseline band
checkout-apiimpacted
payments-svccalls
pg-primary-02root cause
AI Investigatorexplains · never decides

Built on proven, open foundations

OpenTelemetry Apache Cassandra OpenSearch etcd NGINX Apache Parquet Linux · systemd Kubernetes PromQL adapter LogQL adapter OIDC & SAML Ed25519 X25519MLKEM768 ConnectRPC OpenTelemetry Apache Cassandra OpenSearch etcd NGINX Apache Parquet Linux · systemd Kubernetes PromQL adapter LogQL adapter OIDC & SAML Ed25519 X25519MLKEM768 ConnectRPC
Why ObserveX

Full-stack visibility,
without handing over your data.

SaaS monitoring asks you to ship every log line, trace and metric to someone else's cloud. ObserveX runs the whole platform inside your own network, and stays simple to operate.

Your servers, your data

The full platform — intake, storage, analysis and console — runs on your Linux servers or cloud account. Vendor services are outbound-only and never needed at runtime.

data residency: 100% yours

Deterministic answers

Problems come from baselines, anomalies and your live topology, so the same signals always give the same verdict. AI explains what happened and cites the evidence. It never decides on its own.

no language model in the core

Secure by construction

Mutual TLS 1.3 everywhere, a hybrid post-quantum key exchange, signed configuration, one-time enrollment, and masking of personal data at the source.

zero-trust between every component
The platform

One agent. One gateway.
One cluster that understands it all.

Each part is a small, tested module with one job, so the whole platform is easy to reason about, operate and audit.

Shippedobservex-agent

One agent per host

A watchdog with a lightweight core collects host and process metrics, log files and the systemd journal, and receives OpenTelemetry locally. A bounded disk spool keeps data safe through outages.

Shippedobservex-gateway

Unified Gateway

One binary, three modes — environment, cluster or embedded. It relays agents from segmented networks with quotas and backpressure.

Landing nowentities · topology

Live topology, built automatically

Every series and record carries an entity ID. Services, processes and hosts are connected by calls, runs_on and instance_of, with a full System History of what changed and when.

Shipped6 purpose-built stores

Every signal, in the right store

  • MetricStoreCassandra
  • SearchStoreOpenSearch · logs, events, problems
  • TraceStoreWAL + Parquet blocks
  • ConfigurationStoreembedded etcd
  • EntityStorelive graph + history
  • BaselineStoredeterministic, rebuildable
Landing nowbaselines · anomalies

Deterministic Problems

Baselines learn normal behaviour for every entity. Anomalies are correlated over the topology into one Problem with one root cause, not an alert storm.

RoadmapAI Investigator

AI that shows its work

The default model runs inside your network. Every claim links to evidence you can open. Its tools are read-only, and actions always need a person's approval.

0identical cluster nodes by default — no special roles
0purpose-built stores behind stable interfaces
0message buses to run — no Kafka required
1.3TLS everywhere, with mutual authentication
Architecture

From host to answer
in one clean path.

Agents send over mutual TLS, straight to a node or through a Gateway. The NGINX Front Door terminates TLS 1.3 on its own port. Three identical nodes, each with a Node Supervisor and one modular ObserveX Server, do the rest.

HOSTSGATEWAYFRONT DOORMANAGED CLUSTER host-01agent · otlp host-02agent · logs k8s-node-3agent · pods db-07agent · direct Gatewayspool · relay NGINXTLS 1.3 · mTLS:443 / :9443 node-1 leader Supervisor Server self-monitoring metric search trace config entity baseline node-2 member Supervisor Server self-monitoring metric search trace config entity baseline node-3 member Supervisor Server self-monitoring metric search trace config entity baseline
Hover any component to inspect it
No message busBounded, disk-backed buffers in the Agent and Gateway replace Kafka-style infrastructure.
Self-healing nodesThe Node Supervisor handles start order, readiness, restarts with backoff and crash loops.
Local administrationCluster admin is local. Vendor support is outbound-only and off by default.
From signal to answer

How ObserveX turns noise
into one clear Problem.

01 · collect
metrics
logs
traces
events
masked at source card=4111…1111 → [redacted]
P-2041 14 anomalies → 1 Problem · root cause pg-primary-02
AI Investigator · in-network model

"checkout-api p95 latency rose from 180 ms to 920 ms at 14:02. It follows deploy #482 of payments-svc, which doubled its queries per request. Connection-pool waits on pg-primary-02 rose at the same moment."

metric ×3trace ×12log ×41event ×1
Suggested: roll back payments-svc to #481
01

Collect everything, safely

The Agent gathers metrics, logs, traces and events, and masks personal data and secrets before anything leaves the host.

02

Learn what normal looks like

Baselines are calculated for every entity, with seasonality, in plain Go. Same input, same output, every time.

03

Spot real anomalies

Only sustained deviations from the baseline count, so you don't get paged for a one-off blip.

04

Follow the topology to a root cause

Anomalies are joined along calls and runs_on relations into one Problem that names its root cause.

05

Explain, then let a person decide

The AI Investigator writes the story with citations. Remediation is off by default and always needs a person's approval.

Security & trust

Built for teams who
can't afford to guess.

Banks, telecoms, healthcare and government run under strict data rules. ObserveX was designed from the threat model up.

Ask for the threat model
Mutual TLS 1.3, everywhere

Every Agent, Gateway, node and store authenticates with certificates. Identity comes only from the verified certificate.

Post-quantum ready

X25519MLKEM768 hybrid key exchange is preferred, protecting today's traffic against harvest-now, decrypt-later attacks.

Signed configuration

Agent settings are signed with Ed25519 and refused unless the signature verifies.

One-time enrollment

Enrollment tokens are single-use, hashed at rest and claimed atomically. Each Agent gets its own certificate.

Masking at the source

Query parameters, headers, SQL literals and the fields you choose are masked before data leaves the host.

Secret-redacting logs

Every process logs through a redacting handler, backed by leak tests and fuzzing.

Tamper-evident audit roadmap

ProofChain hash-chains every admin change to an external anchor.

Signed supply chain roadmap

cosign signatures and SBOMs on every package, checked before install.

Built for every sector

Wherever data has to stay home.

01

Banking & fintech

Keep transaction telemetry inside your regulated perimeter.

02

Telecom

Observe large, segmented networks through Gateways.

03

Healthcare

Mask patient data at the source, before it leaves a host.

04

Government & defence

Air-gap friendly, with no vendor runtime dependency.

05

SaaS & e-commerce

Follow a checkout from browser to database.

06

Manufacturing & energy

Watch on-premise plants and edge hosts from a single console.

"We needed full-stack visibility, but our regulator would never allow telemetry to leave our data centre. A self-hosted platform built on that model was exactly what we were looking for."
HSHead of SRERegional bank · early customer
"One Problem with a root cause, instead of forty alerts at 3 a.m. That deterministic approach is what won our on-call team over."
PEPlatform Engineering LeadTelecom operator · early customer
"The security model reads like it was written by our own security team: mTLS everywhere, signed configuration, masking at the source."
CSSecurity ArchitectHealthcare network · early customer
Roadmap

Built module by module.
Tested before it ships.

Every module ships with its own tests and an architecture decision record. Here is where we are.

Shipped

Foundation

  • Architecture baseline
  • Signed config & redacting logs
  • mTLS + post-quantum TLS
Shipped

Managed node

  • Node Supervisor
  • NGINX Front Door
  • Self-monitoring
  • Tarball installer
Shipped

Stores & intake

  • Cassandra MetricStore
  • OpenSearch SearchStore
  • Parquet TraceStore
  • etcd configuration
Shipped

Gateway & Agent

  • Unified Gateway (3 modes)
  • Agent: host, process, logs, journal, OTLP
  • Disk spool & forwarding
Shipped

Cluster & entities

  • 3-node cluster
  • Entity model & topology
  • System History
  • Business events
In progress

Problems

  • Baselines & seasonality
  • Anomaly detection
  • Deterministic Problems on topology
Pricing

Priced for your environment,
not a rate card.

ObserveX Managed runs on your own infrastructure, so every installation is different. Our experts size the platform with you and send a quote that fits.

ObserveX Managed · rolling out now
Talk to our experts

Tell us what you need to observe and we'll put together pricing for your setup.

Every installation includes

  • ObserveX Agent for every host
  • Unified Gateway for segmented networks
  • 3-node managed cluster with all six stores
  • Topology, baselines & deterministic Problems
  • mTLS, signed configuration & masking at the source
  • Help from our team with installation

What shapes your quote

  • Number of hosts and services
  • Data volume and retention
  • Cluster size and high availability
  • Support level and response times
Talk to an expert

Let's plan your rollout.

ObserveX Managed is rolling out now. Tell us about your environment and an ObserveX expert will get back to you with pricing, sizing and next steps.

  • Pricing tailored to your hosts and data
  • Guided install on your own infrastructure
  • A live walkthrough for your use case

Prefer email? hello@observex.io

Or explore the demo first

FAQ

Questions,
answered.

Where does ObserveX run?

On your own Linux servers or in your own cloud account. A default cluster is three identical nodes. Each one runs a Node Supervisor, the NGINX Front Door, the ObserveX Server and the stores.

Does any telemetry leave my network?

No. Intake, storage, analysis and the AI model all run inside your environment by default. Vendor-side services are outbound-only, optional and never needed at runtime.

Is the AI making decisions about my systems?

No. Problems are found deterministically from baselines, anomalies and topology. The AI Investigator explains what happened and cites the evidence. Its tools are read-only, and any suggested action needs a person to approve it.

Do I need Kafka or another message bus?

No. ObserveX uses bounded, disk-backed buffers in the Agent and Gateway, so there is no message bus to run.

Does it work with OpenTelemetry?

Yes. The Agent receives OTLP locally, and code-level modules are OpenTelemetry-first. PromQL and LogQL are supported as query adapters on top of ObserveX's own API.

How much does it cost?

ObserveX Managed is priced for your environment, based on hosts, data volume and retention, cluster size and support level. Talk to our experts or email hello@observex.io for a quote.

Can I use it today?

Yes. ObserveX Managed is rolling out now. Our team helps you install it on your own infrastructure. Modules marked "Next" on the roadmap ship as they are completed.

See it working
in two minutes.

The interactive demo walks through every ObserveX module with live, simulated data. No sign-up needed.