Workload identity for AI agents

Identify the runtime carrying an agent without equating that identity to task authority.

01 / The operating case

What changes in a real workflow?

An agent runs in a container with a SPIFFE workload identity and calls a payment API.

Trace the authority boundary
  1. 01Owner
  2. 02Agent + workload
  3. 03Task context
  4. 04Action decision

02 / Decision contract

What the executor must check

Authenticate the workload and bind the running agent and task to it; separately decide payment action and parameters.

InputPrincipal · task · action · target
DecisionGrant ancestry · policy · current state
ResultEnforce · record · verify outcome

03 / Failure and evidence

What goes wrong, and what can be proven?

Failure mode

The service identity is accepted as permission for every agent hosted in the container.

Evidence to retain

Record workload ID, agent instance, task grant, decision and downstream call.

04 / Canonical scope

Why this reference stands alone

Workload versus agent identity compares principals; this page explains workload issuance and binding.

Technical artifact

Bind agent instance to its host

Conceptual trace
host identity: spiffe://org/ns/finance/sa/worker
agent A: task invoice-I42, read only
agent B: task invoice-I43, prepare only
A requests prepare(I43) → DENY
valid host identity is insufficient

Run the denial test

Start two agents on one authenticated worker and swap their task references. Reject the mismatch before the protected API runs.

Primary references

Read the underlying material

Architecture discussion

Choose one consequential action and make its boundary explicit.

Request a Conversation