A new MCP server alters the operational perimeter.
Founding Design Partner Programme · Panama + Bay Area
Help define how real-world AI is authorized to operate.
A limited cohort for organizations and technical partners with real deployments, a willingness to challenge assumptions and the ability to turn learning into operational evidence.
Exploratory first conversation · No certification · No credentials or secrets
Paid professional programme
Participation in the programme is not free. Every accepted collaboration is formalized as a paid professional engagement, with scope, responsibilities and commercial terms agreed before work begins.
Panama
Bay Area
AEGIS
ASSURANCE
Data
Tools
Permissions
Autonomy
Evidence
Human owner
ASSURANCE
Limited cohortSelected for real fit
Concrete deploymentNo deck required
Paid collaborationWritten scope and fee
No certificationThe evidence may say no
01 · Why now
An agent changes faster than the authorization that approved it.
The pressure is no longer only to deploy. It is to prove what was authorized, detect when the system is no longer the same and decide what evidence is required again.
An unused permission is exercised for the first time months later.
A valid authorization no longer describes the real system.
02 · What we are building
From scattered evidence to a decision someone can defend.
Aegis turns a concrete deployment into a verifiable sequence of boundaries, testing, human accountability and revalidation.
Scope
Purpose, owner and the exact system under review.
Boundary
Data, tools, permissions, autonomy and impact.
Challenge
Adversarial hypotheses, failure modes and stop conditions.
Decision
Evidence, residual risk and documented human authorization.
Revalidation
Triggers that invalidate evidence when the system changes.
Design principleAuthorization belongs to a defined configuration, not to the agent's name.
03 · Two bridges, one discipline
Two ecosystems. Two ways to improve the method.
Panama
Operations, adoption and workable governance.
Conversations with organizations and leaders bringing AI into real business processes.
- Operational cases with concrete data, permissions or impact.
- Controls that can be maintained without infinite bureaucracy.
- Human owners, evidence and a real ability to stop.
Bay Area
Architecture, critique and technical depth.
Conversations with builders, AI architects, security engineers, advisors and technical partners.
- MCP, tools, composite permissions and configuration drift.
- Evidence capture without dependence on one vendor.
- Hard questions about what Aegis does not measure and where it may fail.
04 · Selection
We are not looking for an audience. We are looking for counterparts with something real at stake.
Selection prioritizes case quality, access to a counterpart and a willingness to work with uncomfortable evidence.
01Founders and CTOsDirect deployment accountability
02AI ArchitectsTools, permissions and boundaries
03Innovation & OperationsAdoption inside real processes
04Technical partnersIntegration, platform or evaluation
05Security & AssuranceTesting and adversarial critique
06Enterprise AI TeamsGovernance across a growing portfolio
05 · Founding exchange
Early access in exchange for reality, critique and commitment.
The partner brings
- A real deployment or pre-deployment scenario.
- An identified technical or operational counterpart.
- Available evidence within agreed boundaries.
- Direct feedback and a willingness to revisit assumptions.
Aegis brings
- Assurance structure and an operational boundary map.
- Test hypotheses and evidence gaps.
- Priority controls and stop conditions.
- Revalidation triggers when the system changes.
Possible artifactsOutputs that remain useful after the meeting.
Boundary Map
What the system can touch, decide, execute and expose.
Evidence Pack
Tests, limitations and gaps supporting the decision.
Decision Record
Who authorized what, under which conditions and why.
Revalidation Triggers
Changes that require the authorization to be reviewed.
Participation is not free. After the initial conversation, every accepted collaboration is defined as a paid professional engagement, with proposal, scope, deliverables, timeline and commercial terms agreed in writing.
06 · Working model
Five steps. One hypothesis at a time.
Fit
20 minutes to confirm there is a real case and a useful question.
Boundary
Define what system enters, what evidence exists and what remains out.
Scope
Agree timeline, information limits, responsibilities and conditions.
Cycle
Work on the points that could materially change the decision.
Close
Document learning, limits and the next step or a clean exit.
07 · Commercial terms
This is not a free pilot dressed up as collaboration.
The initial 20-minute conversation is exploratory and creates no obligation to proceed. If there is fit, Astrynn provides a proposal proportionate to the case, and work begins only after scope, deliverables, timeline, responsibilities and fee are agreed.
Paid professional programme
- No universal fee.Pricing depends on the system, access, evidence and depth required.
- No open-ended work.Each cycle has a hypothesis, a boundary and a defined exit.
- No promise of approval.The value is in the evidence, including when it requires a stop.
- No automatic public use.Any case, quote or logo requires specific permission.
08 · Non-negotiable boundaries
The programme earns credibility through what it refuses to promise.
It is not a certification.
Participation does not produce an official seal, safety guarantee or regulatory approval.
It does not guarantee Clearance.
The evidence may show that controls or specialist review are still required.
It is not unlimited consulting.
Each collaboration starts with a bounded case, a named counterpart and an agreed scope.
The decision remains human.
Aegis structures evidence and boundaries. The organization retains accountability and authorization.
Nothing becomes a public case by default. Any name, logo, quote or identifiable learning requires prior and specific permission. The first conversation takes place without credentials, secrets, personal data or confidential architecture.
09 · Founder conviction
“Capability without documented authority is not governance.”
Organizations joining now are not being asked to validate a finished idea. They are being invited to help find where it breaks, what evidence is missing and what must be revalidated when the system changes.
Javier Muñoz BurilloFounder, Astrynn Holdings · Creator of Aegis AI Assurance Guardian
10 · First step
Bring a scenario. Leave with a decision on the next step.
The initial conversation is used to determine fit, not to provide a free assessment. If the case should proceed, you will receive a professional proposal before sensitive information is shared or work begins.