# Press Email Template

## Subject Options

- AI agents need a permission layer before they touch the world
- Mission: agents earn permission before they act
- A receipt layer for AI agents doing real work
- The missing approval cockpit for agentic work

## Email

Hi [Name],

I am building Mission: a trust-governed execution layer where agents earn permission to do real work and learn the operator's voice over time.

The angle is not another agent startup. It is the boundary problem created by agentic workflows. Agents can increasingly read, draft, call tools, browse, write code, and touch external systems, but most permission models are still broad allow/block switches.

Mission's Trust Graduation model classifies actions by consequence, prepares approval packets, records receipts, and lets agents graduate only by action class and evidence. The second loop is a compounding voice model: edits, approvals, refusals, and rewrites become durable evidence of how the operator wants work prepared.

The current proof is live here:

- https://gomission.io/proof.html
- https://gomission.io/builders.html

The sharp question is whether this is a real missing layer in the agent ecosystem, or whether it should live inside existing agent platforms.

Would this be interesting for you to look at?

Best,  
Ronen

## Short DM Version

I am building Mission: a permission and voice-learning layer for AI agents. The thesis is that agents should not get blanket tool access; they should earn action-class permission through approval packets, receipts, and evidence, while learning operator voice from edits and approvals. The shortest proof is here: https://gomission.io/trust-demo.html. Press kit: https://gomission.io/press/mission-press-kit.pdf. Would value your critique.