Made in
Europe
Europe
Europe
Deterministic policy enforcement, cryptographic audit trails, and fleet-wide coordination for every agent in your stack. Zero code changes.
Zero code changes. Wrap existing Python, Go, or Node.js agents as-is. The agent runs unchanged as a subprocess; Isonapse intercepts its boundaries from outside. Adopt first, configure later.
The five-stage trust pipeline runs inside the same Docker container you already deploy — wrapped in an .isonapse container. Identity, policy, witness, execution. No new runtime, no new daemon, nothing extra to install.
The early-access beta is open for partners. Join our private Slack to get the CLI, templates, and direct support from the engineers building it.
OIDC federation. Action trace begins, scoped to an authenticated human.
Per-action policy decision: permit, defer to a human, or deny outright.
A signed receipt of the decision, appended to the hashchain. Independently verifiable.
Action runs inside the Secure Agent Container (.SAC) — isolated, identity-scoped.
Outcomes feed local-first memory. Agents remember what worked and improve over time.
Each product solves one layer of the agent infrastructure problem. The CLI is short, lowercase, verb-first — designed to feel native to a terminal.
iso runSecure Agent Container (.SAC). Same agent on Process, Docker, Kubernetes, μVM, or WASM.
iso shieldPer-action policy enforcement. Permit, defer, or deny — deterministically.
iso witnessA hashchain of signed receipts. Regulators verify the chain independently.
iso identityOIDC federation. Every action traces back to an authenticated human via your IdP.
iso memory
Coming soon
Local-first memory and learning. Agents remember what worked, improve over time, and stay inspectable.
Not middleware. Not a decorator a developer can forget to add. Enforcement happens beneath the application — transparent, mandatory, unbypassable.
Every HTTP request, file access, shell command, and DNS query passes transparently through Isonapse's proxy. The policy engine isn't a decorator a developer can forget to add — or an interceptor a misconfigured integration can bypass.
A hashchain of signed receipts. Anchored to a public chain at intervals. A regulator verifies the chain's integrity independently — without access to your systems, your keys, or your network.
Every component degrades gracefully offline. Edge nodes operate autonomously. The same agent container runs on bare metal, in microVMs, or in a browser tab — cloud enhances learning, never correctness.
EU AI Act
Regulatory alignment
Built for secure, auditable execution of AI agents in regulated environments.
Join our private Slack channel to get early access to the beta version.