Run AI agents where your data already lives.
Zybuu builds the infrastructure for AI workloads under your own control — SaaS, private cloud, or fully air-gapped — with the same guarantees at every tier. Abhed, our agent harness, is the first product.
The announcement
Two minutes on why Zybuu exists.
Who we are, the wall enterprises hit, what we build instead, and the first product. Narrated, and made by the same team.
The full series — Abhed, installing it, using it — is shared on request through the access form.
Why we exist
The layer underneath the agent is the one nobody sells.
Every enterprise is going to run agents. The market sells that one way: a vendor's cloud, the vendor's model, your data leaving the building on every turn. A large part of the economy cannot buy it that way.
The problem
Banks, hospitals, defence, government, and anyone with a data-residency clause are offered a choice: wait, accept a weaker product, or build the whole harness themselves. Most are building. Most of it is the same thing, badly.
What Zybuu builds
The infrastructure that lets an organisation run AI workloads under its own control, at whichever tier it operates. Not a chatbot: the part that decides what a model may do, does it safely, records it, and can prove afterwards what happened.
Why Abhed first
An agent harness is that layer for agents. The model is a replaceable part; the harness is what an enterprise actually has to trust — and the one thing a platform team cannot buy off the shelf today.
The scaffold around the model decides more than the model does. A controlled study on SWE-bench Verified found harness-induced variance exceeds model-induced variance by 7.80×, reversing the ranking in 6 of 9 model-pair comparisons. Most of the industry treats that layer as glue around someone else's API. Zybuu treats it as the product.
Products
One product shipped. A platform underneath it.
Abhed Live
A deep agent harness that runs on your own hardware, against a model you host. Twenty model providers behind one abstraction, a tiered execution sandbox, absolute deny rules, a replayable record of every action, and an SDK that embeds all of it without weakening it.
Each shares Abhed's foundations — the same sandbox, policy engine and record — and is stated as what it is: planned, not shipped.
How Abhed competes
We don't compete on the model. We run theirs, or yours.
The incumbents own the developer who can accept their terms — and most regulated firms can, through a cloud region and a no-retention agreement. Abhed is for the ones that cannot: defence and intelligence contractors, sovereign and public-sector deployments, operational networks with no route out, and the banks whose policy is no cloud at all.
Cloud coding agents
- Examples
- Claude Code, OpenAI Codex CLI
- Built for
- One developer, one vendor's models
- Where the model runs
- The vendor's cloud
- Abhed's difference
- Runs where the data is, on any model, with policy and a record the organisation owns.
Terminal harnesses
- Example
- pi.dev
- Built for
- One developer, many models, in a terminal — and better than Abhed for that developer
- Where the model runs
- Wherever you point it
- Abhed's difference
- Built for a platform team serving an organisation: users, tenants, policy, audit, a server, an SDK.
Agent frameworks
- Examples
- CrewAI and frameworks like it
- Built for
- A Python team assembling agents
- Left to you
- The sandbox, the approvals, the audit log, the deployment
- Abhed's difference
- The assembled, hardened thing — with extensions, skills and MCP underneath for the parts that should be yours.
Abhed
- Built for
- Platform teams in places with no route out: defence, sovereign, OT, no-cloud finance
- Where it runs
- Laptop, rack, or air-gapped enclave — one binary
- Model
- Twenty providers, including the one on your GPUs
- What you own
- Every action, every decision, every byte — recorded and replayable.
Descriptions of other products are from their public documentation as of September 2026, and are meant fairly. If we have one wrong, tell us and we will fix it. Open-source harnesses such as OpenHands and Goose also run any model in a sandbox; Abhed's difference from them is the audit record and policy engine in the core, the signed offline bundle, and the SDK that keeps both.
About
Founded on 1 September 2026, by an engineer who has run this problem from the inside.
Zybuu was founded on 1 September 2026 by an AI/ML staff engineer with years spent building and operating machine-learning systems inside large organisations — the kind where the model is the easy part and everything around it is the work: the boundary, the approval, the record, the audit that comes afterwards.
Zybuu is founder-led and building in the open — one product live, the platform beneath it stated as planned rather than promised. Questions, partnerships and access requests all go to support@zybuu.com, where a person reads them.
Evidence
Numbers that survive a technical diligence call.
What is not true yet. Zybuu holds no SOC 2, ISO 27001 or HIPAA certification. There is no support SLA. A human red-team engagement remains outstanding and is not substitutable by the adversarial suite. Those are stated here rather than discovered later, because the buyers this is built for will ask — and an honest answer is worth more than a careful one.
Contact
Building this in the open.
Zybuu is early and founder-led. If you are an investor, a team with data that cannot leave your network, or an engineer who finds the thesis interesting — the fastest way to judge it is to use the product.