Lascaris · Kafka on S3, built in the EU

Kafka on your own S3.

Plus an immutable backup for the Kafka you already run.

Lascaris runs Kafka on the object storage you already own, in any cloud including IBM Cloud, AWS, Azure and Google Cloud, or on-premise and air-gapped. It is Kafka-compatible, so your existing clients keep working. Brokers are stateless and your bucket holds the data, so there are no broker disks to buy and no partitions to rebalance. Agents, policy checks and a signed audit trail are built in. We manage it inside your cloud account, or you run it yourself.

Up to 90% lower streaming storage cost. Your data stays in your bucket, under your keys.
Immutable backup on S3 for Confluent, Apache Kafka, Redpanda and Amazon MSK. Production stays untouched.
EU company. The whole stack runs in your environment, with no vendor control plane. Built for DORA, NIS2 and the AI Act.

Book your call Get the white paper

live decision · your network running

sense

KAFKA TOPICS

orders.events · live stream

0events / s

decide

Your AI agents

reasoning on live state · your models

your modelson-prem

check

Your policy

rule check · passed

enforcedbefore action

record

Audit record

decision #2231 · immutable

lockedAI Act · DORA

inside your networkdata leaving: 0 bytes

Decisions logged

1,284,551

Lock-in

zero

Runs

your cloud

Kafka on S3. Built in the EU.

Streaming on your storage. Agents on the record.

Lascaris runs Kafka on your own object storage, keeps a locked backup of the Kafka you already run, and lets your agents act on live data with every decision signed and kept.

Get the specs

In plain terms

What is Lascaris?

What it is

Kafka on your own object storage. A Kafka-compatible streaming platform with stateless brokers, where every record lives in your S3 bucket. It also keeps an immutable backup of Kafka clusters you already run, and it runs AI agents with policy checks and a signed audit trail on top. AI agents on Kafka: how it works.

Who it is for

Platform and data teams in regulated and sovereignty-minded organisations, including finance, energy, defense, space, health and the public sector. Teams that want the economics of Kafka on S3 without handing their control plane or their metadata to a vendor.

What it replaces

Broker disks, partition rebalancing, and the second production-sized cluster many teams pay for as disaster recovery.

Where it runs

Any Kubernetes with S3-compatible object storage. AWS, Azure, Google Cloud, IBM Cloud, EU sovereign clouds, your own data center, and air-gapped sites.

Why it is different

Built by an EU company. The whole stack, including metadata and the control plane, runs in your environment. Agents, policy checks and a signed audit trail are part of the platform, not an add-on. The core is open source under Apache 2.0 and the segment format is documented, so your data and your exit stay open.

Proof

Open core developed in public under Apache 2.0, with 21 Kafka APIs supported and S3 durability of eleven nines. Built by the original inventors of Apache Wayang.

Your bucket is the cluster.

Lascaris speaks the Kafka protocol and writes every record to object storage you own. Brokers keep nothing on local disk, so there is nothing to copy between them and nothing to rebalance. Your bucket is the source of truth, with eleven nines of durability and retention set by lifecycle rules you control.

Kafka-compatible

Existing clients keep working

Produce, fetch, consumer groups, metadata and admin calls, across 21 Kafka APIs. Point the Kafka clients and tools you already use at one endpoint.

Your storage

Your bucket, your keys

Amazon S3, MinIO, Ceph and other S3-compatible stores, including the S3 APIs of IBM Cloud Object Storage and Google Cloud Storage. Server-side encryption with your own KMS keys.

One endpoint

Clients never see the topology

A Kafka-aware proxy answers every client with one stable address. Brokers scale and rotate behind it, and clients never have to reconnect to a new broker list.

Open format

Read the log without the brokers

Segments use a documented, open format. Processors write Iceberg tables and answer SQL straight from S3, with zero load on the brokers.

A strong fit

Logs, events, ETL feeds and agent context

Workloads where 200 to 500 ms end-to-end latency is fine, and where long retention, low storage cost and full replay matter more than single-digit milliseconds.

Not the right fit today

Sub-10 ms paths and exactly-once

If you need sub-10 ms latency, Kafka transactions across topics or compacted topics, keep those topics on your current cluster. We will help you decide which topics move and which stay.

Elastic by design

Scales on load. No rebalancing.

Lascaris runs on Kubernetes. Its operator sets up a Horizontal Pod Autoscaler for the brokers, so broker pods are added when load rises and removed when it falls. Because brokers keep no data on disk, scaling is a pod change, not a data migration.

Scales on CPU and memory load

By default, Kubernetes adds broker pods when average CPU use passes 70% or memory passes 80%, and removes them when load drops. You set the minimum number of brokers; the autoscaler can grow to four times that.

No rebalancing, ever

Every segment lives in your bucket, and partition ownership is coordinated through the metadata store. A broker that joins reads partition indexes from S3. A broker that leaves takes no data with it. Nothing is copied between brokers.

History never slows the live path

Replays, backfills, Iceberg tables and SQL read segments directly from S3, so heavy historical reads never touch the brokers that serve live traffic.

Kubernetes operator

Clusters and topics are declared as Kubernetes resources. The operator creates the brokers, the autoscaler and the metadata store, and keeps them in the desired state.

Graceful restarts

Brokers are drained through a control API before they restart and give up partition ownership on shutdown, so upgrades roll through without moving data.

One stable endpoint

Clients connect to one address through a Kafka-aware proxy, so brokers can come and go behind it without client changes.

Immutable Kafka backup

A locked copy of the Kafka you already run.

Lascaris continuously copies your topics from Confluent, Apache Kafka, Redpanda or Amazon MSK into an immutable backup on your own S3, under object lock. Your production cluster sees one more read-only consumer and nothing else. Retention is measured in years, at object storage prices, and the backup is readable through the normal Kafka protocol.

Operator error

A topic delete or a bad infrastructure change hits production. Hot replication mirrors the deletion to the standby. The locked backup keeps every segment.

Compromised credentials

An attacker writes bad events or shortens retention with a leaked producer key. The backup runs on separate credentials, keys and bucket that production cannot reach.

No rewind

A bug corrupts events for 47 minutes. Restore the affected topic as of a chosen time into a new topic, validate it, then move consumers over deliberately.

	Hot replication to a standby cluster	Lascaris locked backup
Accidental delete	Replicates to the standby	Untouched under S3 object lock
Leaked credentials	Bad writes replicate too	Separate credentials, keys and bucket
Recovery	Current state only	Topic restore as of a chosen time, into a new topic
Source	Usually the same vendor on both ends	Any Kafka-protocol source
Reference cost, 90 TB	About $12,000 to $38,000 per month for a second cluster	About $2,100 per month in S3 storage
Compliance evidence	Built by your team	DORA Article 12 and NIS2 Annex II evidence pack, restore-test reports
Production impact	Second production-sized cluster to run	One extra read-only consumer

Plan a two-week backup pilot

Is your standby already enough? Read Kafka backup vs replication.

Agents and audit, built in

Agents act on the stream. Every decision is on the record.

Lascaris runs your AI agents next to the data, inside your network. Each action passes a policy check before it runs, and each decision is written to a signed, append-only audit trail on your own object storage. The log your agents read is the same evidence your auditors get.

Boundary · your network

Sense

Agents read live topics, and months of history straight from S3, without load on the brokers that serve production.

Decide

They reason with your own models, running where your data lives, and every proposed action is checked against the policy you set.

Act

Approved actions go out as governed commands, and the decision, its inputs and its outcome are written to the audit trail.

On the record

LASCARIS DECISION LOOP · NODE_AMR_04

REALTIME STREAM

10:42:01.004 SENSE ROS topic /amr_04/sensor_msgs/laser_scan → Kafka on S3 ingestion [120 MB/s, no rebalance]

10:42:01.018 DECIDE Shared memory flagged an obstacle in zone B-12. Agent evaluated an alternative path. Policy check: PASSED (Rule #402: Speed ≤ 0.5 m/s)

10:42:01.025 ACT Published /amr_04/cmd_vel → linear.x=0.4, angular.z=-0.2. Signed audit record committed to S3.

Policy before action

Rules you write are enforced before an agent acts, not reviewed after the fact. A failed check stops the action and is recorded like any other decision.

Signed, append-only audit trail

Every decision is kept with its inputs, policy result and timestamp, on object storage you control and under retention rules you set. Built for the AI Act, DORA and NIS2.

Replay any decision

The full log lives in your bucket, so you can replay what an agent saw and did at any point, for review, testing or after-action analysis.

Deployment

Any cloud. Your account. Your keys.

Lascaris needs Kubernetes and S3-compatible object storage, nothing more. Run it where your data already lives, managed by us inside your cloud account or run by your own team.

AWS, Azure, Google Cloud

Runs on the managed Kubernetes of the cloud you already use, with data in buckets you own.

IBM Cloud

Supported, including a managed backup option operated by Scalytics in IBM Cloud.

EU sovereign clouds

Any EU provider with Kubernetes and S3-compatible storage. Your jurisdiction, your contracts.

On-premise

Your data center and your hardware, with MinIO, Ceph or another S3-compatible store.

Air-gapped

No internet egress required. Images mirror into your own registry and nothing calls home.

No vendor control plane

Data, metadata and control plane all stay in your environment. No outside service can revoke access or change your limits.

Managed in your account

We operate it, you own it

Scalytics runs, patches and monitors Lascaris inside your cloud account, under your keys, from the EU.

Self-run

Your team, our support

Run it yourself on-premise or air-gapped, with verified releases, security hardening and a support SLA under one contract.

Operations Explorer

Mission scenarios in real time.

Explore how Lascaris agents sense, decide, and act across domain-specific operational networks.

Defense: Sensor Fusion & Tactical ISR

Ingests radar streams, optical telemetry, and target tracking vectors directly inside air-gapped enclaves. Agents correlate spatial threats in shared memory, and every tactical recommendation passes a policy check before it is committed to the mission log.

Robotics: Fleet Optimization & Kinematic Safety

Streams multi-robot ROS telemetry (`sensor_msgs`, `nav_msgs`) with zero partition rebalancing delays when connectivity drops. Agents detect path obstructions and execute real-time speed envelope corrections via governed `cmd_vel` topics.

Energy: Substation Load Balancing & OT Security

Processes high-frequency grid telemetry to identify thermal anomalies and unexpected phase shifts. Built-in anomaly detection flags unusual traffic on OT buses while agents re-route load profiles, with every action audited for regulators.

Space: Satellite Telemetry & Orbital Collision Avoidance

Handles bursty downlink telemetry across ground station windows. Agents track orbital drift vectors against live debris catalogs in shared memory, and every maneuver adjustment must pass hard propellant budget guardrails before it runs.

Questions

Common questions.

What is Lascaris?

Lascaris is Kafka on your own object storage. It is a Kafka-compatible streaming platform with stateless brokers that keeps every record in your S3 bucket, an immutable backup for Kafka clusters you already run, and a runtime for AI agents with policy checks and a signed audit trail. It is built by Scalytics, an EU company, and runs fully in your environment.

Will my existing Kafka clients work?

Yes for most workloads. Lascaris supports 21 Kafka APIs, including produce, fetch, metadata and consumer groups, so standard Kafka clients and tools connect to one endpoint. Kafka transactions, idempotent producers and compacted topics are not supported today, so keep those topics on your current cluster and set enable.idempotence=false on producers that write to Lascaris.

What latency should I expect?

Writes are acknowledged after they reach object storage, so end-to-end latency is typically 200 to 500 ms. That suits logs, events, ETL feeds and agent context. For paths that need sub-10 ms latency, keep a low-latency cluster and use Lascaris for the rest.

How is Lascaris different from other diskless Kafka offerings?

The whole stack runs in your environment by default, including metadata and the control plane, with no vendor service in the data or control path. Lascaris is built by an EU company, the core is open source under Apache 2.0, and agents, policy checks and a signed audit trail are part of the platform.

Which clouds and object stores does Lascaris support?

Lascaris runs on any Kubernetes with S3-compatible object storage, including AWS, Azure, Google Cloud, IBM Cloud, EU sovereign clouds and on-premise data centers. Amazon S3, MinIO and Ceph work natively, and IBM Cloud Object Storage and Google Cloud Storage work through their S3-compatible APIs.

Does Lascaris run on IBM Cloud?

Yes. Lascaris runs on IBM Cloud like on any other cloud with Kubernetes and S3-compatible storage, and the managed backup option is operated by Scalytics in IBM Cloud.

Can Lascaris back up Confluent, Redpanda or Amazon MSK?

Yes. Lascaris copies topics from any Kafka-protocol source, including Confluent, Apache Kafka, Redpanda and Amazon MSK, into an immutable backup on your own S3 under object lock. Your production cluster only sees one more read-only consumer.

How does recovery from the backup work?

Recovery is additive. You restore a backed-up topic as of a chosen time into a new topic, validate it, and then move consumers over deliberately. The source cluster is never rolled back, and the backup stays readable through the normal Kafka protocol.

How does Lascaris help with DORA and NIS2?

Backups sit under S3 object lock on separate credentials, keys and storage, restore tests produce timestamped reports, and every state-changing action is logged. The supported distribution includes a DORA Article 12 and NIS2 Annex II evidence pack.

Can Lascaris run on-premise or air-gapped?

Yes. Lascaris runs without internet egress. Images mirror into your own registry, and the only requirements are Kubernetes and an S3-compatible object store inside your network.

Is Lascaris open source?

The core components are open source under Apache 2.0. Lascaris is the supported commercial distribution that integrates, hardens and maintains them as one platform, with a support SLA.

The name

Lascaris

Beneath Valletta, carved into the rock, are the Lascaris War Rooms. From that bunker the air defense of one of the most heavily bombed places of the Second World War was commanded, and the liberation of southern Europe was planned.

Grand Master Lascaris, who led the Order in the 1600s, also built the prototype for the watchtowers that ring the coast, each in sight of the next, relaying a single warning around the island, tower to tower, back to command. A sovereign command center and a trusted signal network. That is what this platform is.

Design Partner Field Notes

Early operator feedback.

Honest assessments from the architects and operators using Lascaris in sovereign environments.

Defense Design Partner

We have been a design partner testing Lascaris as a command system for drone and intelligence operations. Compared to Palantir Gotham it is still missing some of the heavier legacy features. That part is real. But the core ontology modeling, real time telemetry fusion, and analyst workspace already feel solid and highly usable.

What stands out is how responsive the team is. They actually listen and move fast to fix things. Implementation and support have been much smoother than what we experienced with bigger vendors. Running it fully sovereign in our own environment is a big plus.

The platform is evolving fast, and they are clearly determined to build something better for users who want absolute control.

Systems Engineer

Design partner comment on Lascaris from Scalytics.

It lacks some of the legacy modules compared to Gotham, but the basics work flawlessly (ontology, provenance, analyst views). The team is fast and helpful which is refreshing. We set it up completely in our own environment without issues. A much more agile foundation. We use it as command control for intelligence operations in addition to legacy systems. The autonomous agents are the absolute winner here.

Datacenter Operations Partner

As a design partner running renewable energy datacenters, we evaluate platforms on operational reality. Lascaris deliberately skips the legacy bloat. The way it approaches autonomous load balancing and grid coordination is different from older platforms, and honestly feels more thought-through.

Having said that, the foundation is strong and the team is genuinely motivated to close any operational gaps. Their support during implementation was excellent and they stay engaged.

Being able to run the whole stack air gapped and sovereign without the usual vendor dependencies makes it a massive operational advantage.

Operational Architect

We joined as a design partner to test Lascaris in a real operational setting. It does not carry the heavy feature bloat of Palantir, but I think that is exactly the point, since they have a decentralized approach.

On the positive side the platform already delivers excellent fusion capabilities at the edge, and the Scalytics team has been exceptional with implementation and ongoing support. They are hungry to improve it and it shows.

The ability to keep everything sovereign and deployed exactly where we need it is a real advantage for us.

Who builds it

Built by people who have run high-stakes systems.

Sovereign operations rest on trust. Lascaris is built by people who have run systems where every decision had to be accounted for.

Military service

The team are veterans of the German armed forces, served in the Air Force and in ground forces. The experience, accountability and chain of command that sovereign operations require are built in.

Enterprise leadership

The team has held senior and executive roles at E.ON, Allianz, Cloudera, Scout24, and McKinsey, building and running systems in energy, finance, and data at national scale.

Open foundations

Scalytics was founded by the original inventors of Apache Wayang, now an Apache top-level project. A company a co-founder previously built, Ecolytiq, was acquired.

Put your Kafka on your own S3.

Book a call See pricing