The whole platform, in the open
The open source operational data platform.
An operational data platform you can read, run, and own, built on standard parts and shipped under AGPL-3.0.
Released under AGPL-3.0 · Self-host for freeAGPL-3.0
Self-host, fork, audit, run it air-gapped.
10×
Routinely a tenth of the cost of proprietary systems.
8 parts
Postgres · Pulsar · ClickHouse · Neo4j · Keycloak · Kvrocks · Valkey · Vault. Standard, escapable.
What's in the box
Eight standard parts. Novel composition.
We don't reinvent storage, streaming, or graph. We compose the open-source infrastructure your team already runs. The substrate is yours to read, swap, or replace.
Neo4j
Asset and process relationships, ontology, lineage edges. Your digital map of your business.
ClickHouse
Columnar scans over billions of signal rows per second. Advanced compression with Gorilla and ZSTD for compression ratios up to 1000×.
Apache Pulsar
Distributed, durable, multi-tenant pub/sub for live ingestion. Can handle billions of data points per second.
PostgreSQL
Metadata, config, transactions and the things that need to be boring.
Apache Kvrocks
Persistent key-value store for event key mappings. RocksDB on disk, Redis protocol on the wire.
Valkey
Session state and hot lookups, held outside the services so any instance can answer any request.
HashiCorp Vault
Credentials and connection details issued at runtime, so nothing sensitive sits in a config file.
Keycloak
OAuth2 and OIDC. Point it at your own identity provider — users, roles and tokens stay yours.
Why the model wins
Open source is how infrastructure becomes the standard.
The platforms running the world's infrastructure did not get there by being sold harder. They got there by being the thing an engineer could pick up, run and build on without asking anyone, and a standard set that way is very hard to displace.
The adoption loop
-
01
Zero distribution friction
Anyone can run it tonight.
No licence request, no evaluation agreement, no call to book. The code is there to download, run locally and deploy, and that is what turns the people who try it into the people who ask for it.
-
02
Bottom-up adoption
The people who use it decide first.
Developers and operators test it against their own data and form a view long before procurement hears the name. Nobody has to speak to us for that to happen, which is rather the point.
-
03
Faster buy-in
The evaluation happens before the meeting.
Technical teams champion what they have already run. By the time this reaches a budget conversation it has usually been installed, tested and argued for by the people who will operate it, so the hardest part of the decision is already done.
-
04
How a standard forms
The default is picked before it is bought.
Linux and Terraform did not win procurement, they won adoption. By the time a contract existed the decision was made, and everything after that has to argue against a default already in the building.
and the next team already knows it
What the loop pays for
Growth comes from people already running it.
Moving a team that already runs the platform onto a paid plan costs a fraction of winning a name that has never heard of us. That difference shows up in the price.
Security patches are somebody's job.
Self-hosting is real work: patches, scaling, database migrations. Most teams would rather spend that time on their own operation than on ours, and that is what the managed tier is for.
More eyes, and a faster clock.
The people reading, patching and extending the platform are not limited to the people we can hire, so work lands as a community ships it rather than when a roadmap allows. The component that becomes the standard is also the one the market trains people on, so you hire the skill instead of buying it back.
Agents read source, not brochures.
Public code is what these models were trained on, so an assistant already knows the components this platform is built from. Working against your data, an agent can read how it is stored, modelled and served instead of inferring any of it, so it can check itself against the source rather than guess. Why that gap gets more expensive every year.
AGPL-3.0, in plain English
What this actually means for you.
You can read every line.
No proprietary blobs, no closed runtime. The entire stack is browsable on GitHub.
You can self-host, forever.
For free. No seat caps, no host caps, no telemetry tied to a paid plan.
You can fork it.
If we ever disappear, your team can keep the platform alive on its own terms.
Your own code stays yours.
The platform is AGPL-3.0, but the SDK clients are Apache-2.0. Build them into your own software, proprietary or not, without the copyleft following them in.
For your team
Four reasons engineering-led teams choose this path.
None of these is a promise. Each one is something your own engineers can check before anybody talks to us.
Yours to run, read, audit.
AGPL-3.0. Fork it, deploy it air-gapped, change a line of code on a Friday afternoon. The vendor-risk conversation is over before it starts.
Components your team already knows.
Postgres, Apache Pulsar, ClickHouse, Neo4j. Operable by any competent platform team. No proprietary storage to escape from later.
A global build, not a black box.
A platform built and maintained in the open, by people whose work you can read. Issues, pull requests, and decisions are public artifacts.
Priced like infrastructure.
Not a discount, a different cost structure. Engineers find it, run it and decide before we ever speak, so what you pay covers the platform and the people who keep it running.
The team behind it
More than a platform, built to grow with you.
IntelliStream's open-source platform is designed for accessibility. It breaks down the barriers between technical and non-technical teams and enables seamless collaboration.
Open by default
Source available under AGPL-3.0. Every commit, every issue, every release in public. Read the code, audit it, run it on your own infrastructure, and fork it if we ever disappear.
Standard components
PostgreSQL, Apache Pulsar, ClickHouse, Neo4j. No proprietary storage you'll never escape, no novel runtime nobody's heard of. Operate it like you operate the rest of your stack.
Data enthusiasts, shipping.
A team of data enthusiasts who love this problem. When you have a question, you reach someone who can help. We answer our own email.
Where this comes from
I learned this craft in the open.
I didn't learn distributed systems from a vendor manual. I learned them by reading other people's code: storage engines, consensus protocols, and query planners that a global community had worked out in public, free to learn from. That's what open source really is: knowledge sharing at a scale no single company could ever match.
That knowledge compounds. Once you understand the parts, you can compose them into systems that solve real business problems in finance and oil & gas, at a fraction of the closed-source price tag. Not by cutting corners, but because the hardest engineering was already done, in the open, by the community.
HashiCorp, Red Hat, and Canonical proved you can build a serious company without closing the knowledge loop. IntelliStream follows that lineage. The platform is AGPL-3.0 because the knowledge that built it was given to us freely. Passing it on is the only honest way to build on it.
The hardest problems in computer science are solved in public. Open source let me learn them, and turn them into solutions industries can actually afford.
Get started
Two commands. Ten minutes. No sales call.
A working instance with sample data, running on your own machine. Self-host for free under AGPL-3.0. When you have a question, you reach a person.
No card · No discovery call · AGPL-3.0