Runs in your VPC, against your database. Every action attributable, every risky one gated, every run traced. This is what production ready AI agents actually require.
Every request carries an identity derived from a verified credential, never the request body.
Run the self-hostable runtime inside your own environment, or take managed hosting. Either way, conversations, user data and credentials stay inside the boundary you deploy into.
Organizations, users, applications, MCP servers, databases, secrets and environments are isolated by tenant, enforced with per-user credential encryption.
Encrypted credential vaults, API-key issuance, rotation and revocation, and AES-256-GCM encryption with per-user keys, so every action is attributable to an approved identity.
Usage is metered at the execution level, with per-tenant limits and usage-limit guardrails so one workload cannot exhaust shared capacity.
The question is never whether an agent can act. It is whether you can explain what it did.
A policy mechanism intercepts work at five seams and can monitor, redact, block, escalate or require human approval.
Shipped with the platform and enabled per application by policy. No bespoke safety layer to write.
Named actions hold at an approval gate before execution: refunds, deletions, transfers, publishes.
Structured logs, OpenTelemetry traces, per-call latency decomposition, token and cache usage, retrievable execution histories.
Deepening the governance surface beyond tenant isolation and per-user credential encryption.
Re-run a workflow against the same conditions, and attribute spend to applications, teams and individual runs.
Enable Verifiable Execution and selected records are anchored to a cryptographic ledger, proof a counterparty can check without trusting the operator. The rest of the deployment is unchanged.
Self-host the runtime in your own environment, or let us host it. Same platform, same controls.
| Deployment model | Primary value | Status |
|---|---|---|
| Business workspace | Natural-language creation, managed integrations, shared policies. | Now |
| Enterprise deployment | Dedicated environments, tenant isolation, approvals, observability. | Now |
| Infrastructure integration | CLI and SDK that embed Kodeus operating services inside other products. | Now |
| Public ecosystem | Developer-built applications, capability marketplace, usage-based monetization. | Maturing |
Bring a workflow that matters and we will walk through the identity model, governance boundary and deployment shape with your team.
