Resolve identity once; let every engine consume it¶
October 4, 2026 · Kishore Babu Gutta
Authentication often becomes fragmented when each execution transport implements its own login story.
REST obtains a bearer token.
A CLI reads a native credential store.
A Kubernetes workload uses workload identity.
CI exchanges OIDC claims.
If each engine independently decides identity, the product eventually has several subtly different security models.
Engineering Runtime uses a different direction. The description below is the shared-authority design, not a claim that every credential works across every transport.
Resolve before transport¶
Execution surface
↓
Identity + context resolution
↓
Admitted runtime session
↓
REST / GraphQL / Command / File
For supported provider operations and authentication sources, Runtime resolves the admitted identity and target context before dispatch.
Each execution engine then consumes authority in the form its transport supports. For example, in v0.13.0 an ENGRT GitHub session and native gh login have independent lifecycles; authorization cannot be assumed to transfer automatically between them.
For example:
- an API engine may consume an access token,
- a command engine may rely on the corresponding native authenticated session,
- a workload may rely on cloud-native identity,
- a local developer may use an existing provider login.
The goal is a shared, inspectable authority decision where the provider and transport support it, with differences made explicit.
Why this is useful¶
It gives the Runtime one place to reason about:
- identity precedence,
- context,
- interactive versus non-interactive surfaces,
- short-lived credentials,
- fallback behaviour,
- evidence.
A transport change should not silently change the security model; any unsupported authentication parity must be reported honestly.
Authenticate using the platform. Govern using the Runtime.
I think of this as Transport-Independent Authentication: resolve identity and context once, then let execution engines consume the result.