Gateway decides, never executes on your hosts
The Gateway admits transactions, manages PKI, and coordinates deliberation. It publishes each envelope to one exact session channel. There is no broadcast.
Verify, then execute.
We build ztd, a zero-trust execution layer for AI agents and the infrastructure they touch. AI systems propose actions; they never get direct control of real systems. The model can ask. The machine that owns the workload decides.
// ztd is Zero Trust Data: typed intent, independent verification, governed execution, and local evidence.
01 / The problem
Most AI agent stacks collapse four separate jobs into a single process. That is a dangerous mix. ztd pulls them apart so the thing that reasons is never the thing that authorizes or executes.
→ one process, one trust boundary. Whatever the model decides is what runs, and the record of it lives in the same place.
This is not a sandbox. It is a governance boundary.
02 / How ztd works
A client or agent submits a typed intent. The Gateway validates policy, applies human or quorum approvals when needed, and sends the action over an outbound-only connection to a local Operator. The Operator verifies the request on its own, executes only the approved action, and writes its own signed evidence.
MCP, A2A, governed HTTP dispatch, direct envelope, or CLI.
expresses intentAuthenticates over mTLS, builds a canonical GovernanceEnvelope, runs L1–L3, binds it to one Operator session.
ztd gw startNo inbound listeners. Pulls from its session channel, re-verifies everything in L4, executes in L5.
ztd operator startEXECUTING and COMPLETED/FAILED receipts, plus a hash-chained commitment ledger on the host.
local-first auditThe Gateway admits transactions, manages PKI, and coordinates deliberation. It publishes each envelope to one exact session channel. There is no broadcast.
Each Operator recomputes the transaction hash, checks the replay nonce, expiry, and state root, and re-runs doctrine locally. It doesn't take the Gateway's word for it.
The Actuator signs a receipt before and after every action and appends to a hash-chained ledger. The host's local evidence is authoritative; Gateway copies are only mirrors.
Under human-approval postures, mutating actions suspend until someone approves with a WebAuthn passkey or the enrolled CLI certificate (ztd auth approve --cli). Read-only actions don't need that step.
L2 Consensus requires K-of-N Ed25519 signatures from enrolled member keys over the transaction hash and decision.
L1 Doctrine screens every envelope for forbidden patterns and ATT&CK-style heuristics in every posture, on the Gateway and again on the Operator.
03 / The interlock
Every operation on a governed path crosses the same five-layer pipeline. A rule violation, or a proof the posture requires but the transaction lacks, rejects it.
Forbidden-pattern matching, MITRE ATT&CK heuristics, and command safety analysis for things like reverse shells, privilege escalation, destructive disk operations, and credential theft.
OwnerGateway + OperatorK-of-N multi-signature Ed25519 authorization from enrolled member keys, each signing the transaction hash and its decision.
OwnerGateway + OperatorHuman authorization for mutations: the transaction suspends until a person approves it with a WebAuthn passkey or, via ztd auth approve --cli, the enrolled CLI certificate. An outbound Operator re-verifies the recorded approval and its Ed25519 signature over the transaction hash.
Pre-dispatch gate on the target host: durable nonce reservation, expiry, typed payload decode, local doctrine, hash recomputation, state Merkle root, and posture-required proofs.
OwnerExecuting OperatorThe single execution boundary. Signs an EXECUTING receipt, appends a signed commitment, mints a short-lived capability bound to the transaction hash, runs the typed handler, dissolves the capability, and signs the final receipt.
OwnerExecuting Operator| Posture | L1 | L2 | L3 |
|---|---|---|---|
| doctrine | enforced | audited | audited |
| consensus | enforced | enforced | audited |
| ratify | enforced | audited | mutations |
| notary | enforced | enforced | mutations |
04 / Design lineage
The 1972 reference monitor said every access should pass a mediator that is always invoked, tamper resistant, and small enough to verify. Replace subjects and objects with agents and actions, and you get the shape ztd is built around.
every governed path → five layers
Every operation entering a governed path goes through L1–L5, and the L5 Actuator is the only place execution happens. Operators take work from one session channel and expose nothing inbound.
signed receipts · hash-chained ledger
Receipts are signed at the execution site, commitments chain together in a local ledger, nonces are reserved durably to stop replays, and sensitive results are encrypted at rest with AES-256-GCM.
independent re-check · public source
The executing Operator recomputes hashes and re-checks proofs without trusting the Gateway. The code is published and the documented invariants carry stable IDs, so you can audit claims against the source.
05 / Zero trust, all the way down
ztd auth approve --cli.06 / Local-first, air-gap ready
Under ztd's Local-First Audit Architecture, the host where an action runs is the authoritative record of what happened there.
ztd binary runs as the Gateway and as each Operator.POST /mcpModel Context Protocol over HTTP, or stdio via ztd mcp stdio
POST /api/v1/a2a/callAgent-to-agent skill calls
POST /api/v1/operators/commandsGoverned HTTP dispatch for enrolled apps
POST /api/v1/operators/intentsPlain-language intent; the Gateway Tribunal authors typed run_shell_command arguments that pass through MCP_CALL governance
POST /api/v1/governance/envelopesPre-built envelopes; missing proofs are never synthesized
ztd operator runCLI dispatch to bound Operators
32 typed native tools in the reference binary: database triage, logs, processes, containers, filesystem, network and TLS, Git, cloud/Kubernetes inspection, denylist-screened shell commands, Operator deployment, and audit receipt queries.
07 / Quick start
Install the signed ztd binary, start a Gateway, sign in, and run a first governed action. Safe to rerun. Evaluation and other non-production use are free under BSL 1.1. Production use needs a commercial license plan below.
No root, no admin, no toolchain. Checksums verified before install. Binaries and Sigstore bundles: Releases.
./ztd tuiWatch live governance envelopes./ztd login webOpen the console# install, start a Gateway, sign in, run a first governed action $ curl -fsSL https://zerotrustdata.io/start.sh | sh # binary only: curl -fsSL https://zerotrustdata.io/ztd.sh | sh
# install, start a Gateway, sign in, run a first governed action PS> irm https://zerotrustdata.io/start.ps1 | iex # binary only: irm https://zerotrustdata.io/ztd.ps1 | iex
$ git clone https://github.com/zerotrustdata/ztd.git $ cd ztd $ bash scripts/linux-setup.sh --build-only $ make quickstart # macOS: bash scripts/macos-setup.sh --build-only # Windows (PowerShell 7): pwsh scripts/windows-setup.ps1 --build-only
# sign in as the owner with a passkey, then approve workloads $ ./ztd login web $ ./ztd auth enroll pending # set REQUEST_ID to a request ID listed by pending $ ./ztd auth enroll approve "$REQUEST_ID" --yes # headless: ./ztd login for a CLI-only mTLS identity
# Gateway + embedded Operator with data, inference, and observer capabilities # set OLLAMA_HOST to your Ollama host name or address $ make gw GW_ARGS="--inference-ollama-endpoint http://${OLLAMA_HOST}:11434" $ make status # then sign in at https://localhost:8443/console/
08 / Pricing
Your engineers are already connecting Claude Code, MCP servers, and autonomous agents to internal systems. Lock in production licensing on a corporate card today and receive a License Certificate for Finance and your auditors within one business day. Roll out the Operators at your own pace. Every plan runs the same audited binary, so you can verify every claim before you buy.
For engineering teams taking the source-available binary to production with auditor-ready cryptographic evidence.
Lock in production licensing today and deploy across 50 hosts at your own pace.
Full-scale deployment licensing for regulated infrastructure, hybrid cloud, and disconnected / DDIL environments.
09 / OEM license
Don't burn 9 months of engineering runway building SPIFFE PKI, mTLS session routing, WebAuthn passkey step-up, and hash-chained local SQLite ledgers from scratch. Exempt your product from BSL 1.1 restrictions and ship our static Go governance engine inside your SaaS, appliance, or federal deliverable.
ztd inside proprietary closed-source products and managed services10 / Architecture sprints
ztd and OpenDevOps.ai were built by Danny Barbour—a 30-year Principal Infrastructure, Storage & SRE Architect who built Igneous SRE 0→1 through its Rubrik acquisition and IPO (NYSE: RBRK), commanded Fortune 500 storage outage bridges at Quantum, and commissioned U.S. Navy warship IT from bare steel. Skip the 6-month executive search and book a fixed-price technical sprint.
A brutal, senior-level technical audit of your AI agent boundary, Kubernetes/cloud platform, or petabyte storage pipeline.
Drop a Principal Forward-Deployed Architect directly into your repo and cloud environment to wire and ship the solution.
ztd L1–L5 governance OR autonomous SRE remediation on AWS, GCP, Kubernetes, or on-premOn-demand access to a 30-year Principal SRE, Storage & Security Architect when your team hits a wall in production.
ztd Fleet Pack active for the duration of your retainer11 / Track record
ztd wasn't built by prompt engineers guessing how production works. It was architected from the ground up in static Go by a Principal SRE and Storage Architect who has commissioned warship IT from bare steel, operated Nike's global backup infrastructure for 15 years, and scaled multi-petabyte SaaS reliability from 0-to-1 through acquisition and public IPO.
./ztd eval conformance run, the governance conformance suite, to check every L1–L5 claim locally. It needs no model.// end of transmission
Clone the repo to test locally, lock in a 50-Host Fleet Pack on a corporate card, or book a fixed-price Principal Architecture Sprint. Email Danny directly.
12 / Contact
Send a quick email with what you're building, which license pack you need, or what production fire you need fixed. You'll get a direct written response from Danny Barbour within one business day.