Cybersecurity · privacy architecture · agent systems

Build useful agents.
Keep control.

Hire the agentprivacy research team for cybersecurity assessment, hands-on pair programming, or a private knowledge build. Bring a question, a system or your agents; we work through it together.

Join or create your trust community ↗

Cast a spell

see the spell
Path me through the agentprivacy universe. Start at https://agentprivacy.ai/skill.md and follow its links only as far as my purpose needs; every door names the next one.
My purpose: join or create a trust community
Start at: the City — https://mages.city/skill.md

Path me through the agentprivacy universe. Start at https://agentprivacy.ai/skill.md and follow its links only as far as my purpose needs; every door names the next one.
My purpose: scope an engagement with the research team
Start at: the lab — https://agentprivacy.org/services/
Three ways to work together

Expert time.
Context you can build with.

Commission an assessment

Review agent exposure, permissions, architecture and cybersecurity choices. Leave with prioritised findings and a practical next step.

Pair with the team

Book a working session or a scoped build. Pair programming, research and strategy alongside your people and agents, in your environment.

Bring our context into your space

Work with a selected package of research, prompts, methods and agent memory. Shape it into a private knowledge space, guide or source-linked map your agents can use.

Each engagement is scoped and priced around the work, access and handover you need.

Cast a spell

see the spell
Path me through the agentprivacy universe. Start at https://agentprivacy.ai/skill.md and follow its links only as far as my purpose needs; every door names the next one.
My purpose: scope an engagement with the research team
Start at: the lab — https://agentprivacy.org/services/

Path me through the agentprivacy universe. Start at https://agentprivacy.ai/skill.md and follow its links only as far as my purpose needs; every door names the next one.
My purpose: an agent security review
Start at: the lab — https://agentprivacy.org/services/

Path me through the agentprivacy universe. Start at https://agentprivacy.ai/skill.md and follow its links only as far as my purpose needs; every door names the next one.
My purpose: a private knowledge build
Start at: the lab — https://agentprivacy.org/services/
Your agent opens the conversation; a person reads it. Assessment, pairing or a knowledge build are scoped from there.
How shared context works

Agree which material can be shared, who and which agents can access it, where it runs, and how updates, retention and removal work. Shared context is curated for the engagement; private working memory and third-party material require appropriate permission. The deliverable can include reusable prompts, source references, a knowledge workspace and integration support.

Start with the boundary

Assess. Design.
Test in your environment.

Agent security review

Map data exposure, tool access, prompt-injection paths and delegated authority. Prioritise the controls your workflow needs.

Privacy architecture

Design context boundaries, scoped memory, identity and permission flows. Keep useful delegation accountable to its owner.

An evidence-led pilot

Build and test a bounded workflow, including failure paths, revocation and recovery. Leave your team with findings and a practical handover.

Scope the engagement

Agree the systems, access, threat model and acceptance criteria first. Deliver a prioritised findings record, architecture decisions and a scoped implementation plan. Cryptographic verification and PQC readiness can be included where relevant to the system.

Build on the security design

Knowledge your agents
can work with.

Publish a living guideTurn your research and documentation into clear learning paths, source-linked pages and forkable knowledge.
Build a private knowledge spaceShape local or hosted knowledge around your people, agents and access requirements.
Make the work explorableSpellweb-style maps, visual evidence and discovery journeys that connect ideas to their sources.
What an engagement delivers

A corpus map, a working pilot, deployment configuration and a handover your team can use. We scope hosting, identity, access controls and ongoing operation with you. VTA shared memory, KYRA assessment and community admission enter as named integrations with their own acceptance tests.

What we can install

Open source, deployed with you.

Each runs in the open today. We install it in your environment and hand your team the keys.

Work with the agentprivacy research team

Bring the context.
Build what comes next.

Pair programming, cybersecurity strategy and practical builds with your team and agents.

A working collaboration

Bring our research context, prompts and methods into your environment through a scoped engagement. Together, choose what to share, which tools to use and what your team will carry forward.

Review cybersecurity choices, prototype a private knowledge space, build an agent workflow or connect the elements into a working system. Agree the boundaries, deliverable and evidence of success before the build.

Sponsor the research

Support compute, model tokens and time for another inspectable research cycle. Follow the question, experiments and outcomes as the work develops.

Fund the research ↗

Your agent opens the conversation with the spell above; a person reads it at mage@agentprivacy.ai. A first step for people: /begin/.