Every Zero Trust architect already understands the protect surface model. What eats the afternoon is the platform — clicking through screens to build what was designed an hour ago on paper. The AUXO MCP Server, open-source from ON2IT, closes that specific gap: protect surfaces, transaction flows, measure assignment, and incident cases become natural language operations inside Claude, VS Code Copilot, and any MCP-compatible client. No portal. No context switch.
Domain · Zero Trust
Protect Surface Management
Create, update, and manage protect surfaces, locations, states, assets, contacts, security measures, and transaction flows, the full operational Zero Trust lifecycle, through conversational AI interaction or automated toolchain integration.
Domain · Case Management
Incident & Ticket Handling
Create, update, escalate, and close support cases and security tickets directly from the AI assistant context, without switching tools during active incident response. Analysts stay in their workflow. Case management happens inside it.
In Brief
What it is
An open source MCP (Model Context Protocol) server that exposes AUXO’s Zero Trust and Case Management APIs as callable tools to any compatible AI assistant. The AI does not just advise on Zero Trust, it operates it. Protect Surfaces get created, flows get mapped, measures get assigned, and cases get opened, all from natural language inside the tools architects and analysts already use.
Why it matters
Zero Trust breaks down in the gap between intent and execution, not in the theory. Architects design Protect Surfaces, then open a UI to build them. Analysts spot an anomaly, then switch tools to log it. The MCP server removes the switch. The AI assistant becomes the operational interface to AUXO, faster, more consistent, and scriptable.
How it deploys
Three paths: a Claude Desktop extension (double-click install), pre-built binaries for macOS, Linux, and Windows, or a container image from the GitHub Container Registry. Domains auto-enable based on which API tokens you provide, Zero Trust, Case Management, or both. No code required to start. Go 1.24 or newer only if you are building from source.
What it changes
For architects, Protect Surface definition and flow mapping stop being a document-then-platform relay: list existing surfaces, create new ones from the conversation, assign measures from the AUXO catalogue, all in one pass. For SOC analysts, case management becomes a single-context operation: detect, document, escalate, and track without leaving the investigation thread.
"This is what Zero Trust operationalization looks like when the interface catches up with the architecture. The AUXO MCP Server puts the full Zero Trust lifecycle, and the incident response loop, inside the AI toolchain your team is already using. No portal. No context switch. No gap between intent and execution."ON2IT
The practitioner problem this solves
“Every Zero Trust architect has experienced the same friction: you understand the Protect Surface model, you know what flows need to be mapped, and then you open the platform and spend the next hour navigating UI instead of doing architecture. The MCP server inverts that — the AI navigates the platform. You do the architecture.”
An AI that acts, not just answers
The Model Context Protocol (MCP) is an open standard that lets an AI assistant call external tools, APIs, and services as native operations inside a conversation. Configure an MCP server and the assistant stops reasoning from a distance: it calls the tool, gets the result, and works from what is actually true right now.
For AUXO, that interaction is real today:
# Architect asks Claude: "List all current protect surfaces and identify any that have no transaction flows mapped yet." # Claude calls: listProtectSurfaces() -> returns current surfaces listTransactionFlows(protectSurfaceId) -> per surface # Claude responds with a structured gap analysis, then asks: "Should I create flows for the three surfaces with no mappings?" # Architect confirms. Claude calls: createTransactionFlow(source="CRM", target="DataWarehouse", ...) createExternalFlow(surface="CustomerPortal", direction="inbound", ...)
This isn't Q&A. It reads state, identifies gaps, proposes actions, and executes on confirmation. The architect stays in control of every decision. The AI removes the friction of carrying it out.
"The difference between an AI that tells you what to do and an AI that does it with you is the difference between a consultant and a co-pilot. MCP makes AUXO the latter."
Two tokens, twenty-seven tools
Zero Trust and Case Management are independent domains, each behind its own API token. Provide the Zero Trust token and those tools activate. Skip the Case Management token and those tools simply don't appear, no unused configuration, no domain you didn't ask for.
| Tool | Domain | What it enables in practice |
|---|---|---|
createProtectSurfacelistProtectSurfacesupdateProtectSurfacedeleteProtectSurface |
Zero Trust | Full Protect Surface lifecycle: create new surfaces from architecture discussions, list and audit existing ones, update definitions as the environment changes. The AI can propose Protect Surfaces based on a described environment and create them directly on confirmation. |
createTransactionFlowcreateExternalFlowlistTransactionFlowsdeleteTransactionFlowdeleteExternalFlow |
Zero Trust | Transaction flow mapping, the step most implementations treat as a manual documentation exercise. The AI can interrogate existing flows, identify unmapped surfaces, propose flows based on described data paths, and create them. External flows (to or from outside the organization) are handled as a distinct operation. |
listMeasureslistProtectSurfaceMeasuresupdateProtectSurfaceMeasureremoveMeasureFromProtectSurface |
Zero Trust | Security measure assignment from the AUXO catalogue. The AI can search available measures by keyword or context, list what is already assigned to a Protect Surface, update implementation status, and remove measures that are no longer applicable, making the measure lifecycle conversational rather than navigational. |
createLocationlistLocationsupdateLocationdeleteLocationcreateStatelistStateslistContactslistAssets |
Zero Trust | Supporting Zero Trust data structures: locations, states, contacts, and assets. Listing and managing these through the AI assistant means environment context is always available in the conversation without requiring platform navigation. |
createCasegetCaseupdateCasePriorityupdateCaseSubjectescalateCasedeescalateCaseaddNoteToCasecloseCase |
Case Management | The full incident case lifecycle inside the AI assistant. An analyst detecting an anomaly can create a case, document findings as notes, escalate based on developing severity, and close when resolved, without leaving the investigation context. The AI assistant becomes the single interface for both analysis and case management. |
Practitioner signal
“Twenty-seven tools. Two tokens. The entire operational Zero Trust and incident case lifecycle available to any AI assistant that supports MCP — without writing a line of integration code.”
Running in two minutes, or running at team scale
One architect on Claude Desktop and a platform team running a fleet of containers get the exact same capability, not a scaled-down version of it. Pick the path that matches how your team already works.
Path 01 · Simplest
Claude Desktop extension. Download auxo-mcp-server.mcpb from the releases page. Double-click it, or drag it into Claude Desktop settings. Claude prompts for API tokens during setup. No configuration files. No terminal. Running in under two minutes.
Path 02 · Flexible
Pre-built binary. Platform binaries for macOS ARM and Intel, Linux AMD64 and ARM64, and Windows. Place it on your PATH or reference it directly in an MCP client configuration. Works with VS Code Copilot, Claude Desktop manual config, and any other MCP-compatible client. Tokens are passed as environment variables.
Path 03 · Scalable
Container / HTTP mode. Published to the GitHub Container Registry. Run in HTTP mode on port 8080 with tokens as environment variables. Clients connect to the /sse endpoint. Place it behind a reverse proxy with TLS for production. Suitable for shared team environments and CI/CD integration.
{
"servers": {
"auxo": {
"type": "stdio",
"command": "path/to/auxo-mcp-server",
"env": {
"AUXO_ZT_TOKEN": "your-zero-trust-token",
"AUXO_TICKET_TOKEN": "your-ticket-token"
}
}
}
}
Tokens are requested by contacting ON2IT support. Each domain requires its own token, teams that only need Zero Trust operations request only the Zero Trust token. The server auto-detects which domains to enable based on which tokens are present.
"The installation surface is deliberately minimal. The Claude Desktop path requires no terminal, no config file, and no Go installation. The complexity scales with the deployment requirement, not the other way around."
Three workflows every AUXO team runs slowly today
Not theory, the three workflows every Zero Trust implementation team and every SOC running on AUXO repeats constantly, and repeats slowly.
Workflow 01 · Protect Surface definition and gap analysis
Before
The architect describes the environment in a design document, maps DAAS elements manually, navigates AUXO to create Protect Surfaces one by one, and cross-references transaction flows in a separate pass.
With MCP
The architect describes the environment to the AI assistant. It queries existing Protect Surfaces, identifies gaps, proposes new surfaces based on the described DAAS elements, and creates them on confirmation. Transaction flows are mapped in the same conversation, the design document and the platform state are built simultaneously, not sequentially.
The AI can also run a continuous gap check: "List all Protect Surfaces with no assigned measures from the AUXO catalogue and suggest the three most relevant for each." This takes the catalogue search, previously a manual cross-reference task, and makes it a single query.
Workflow 02 · Security measure assignment and implementation tracking
Before
The analyst searches the AUXO measure catalogue for relevant controls, assigns them to Protect Surfaces, tracks implementation status in a separate pass, and repeats for each surface in scope.
With MCP
The AI searches the catalogue by context, for example "Find measures relevant to protecting an on-premises ERP system with external API access", returns ranked results, assigns selected measures to the Protect Surface, and sets initial implementation status.
Updates to status as controls are implemented are a single natural language operation: "Mark the network segmentation measure on the ERP surface as implemented." For teams conducting regular Zero Trust maturity reviews, the AI can generate a current-state summary across all Protect Surfaces and measures in a single query, replacing what was previously a manual export and analysis exercise.
Workflow 03 · Incident investigation and case management in a single context
Before
The SOC analyst detects an anomaly in AUXO, opens a browser tab to the case management system, creates a case manually, adds investigation notes by copying from the detection context, escalates through a separate workflow, and returns to the investigation.
With MCP
The analyst works in the AI assistant throughout. Detection findings, analysis, case creation, note addition, escalation, and closure all happen in the same conversation thread. The AI assistant holds the full context of the investigation and applies it directly to case documentation.
What the analyst observed and what the case record contains are the same thing, without transcription. During active incidents, this is not a convenience improvement. It is a speed and accuracy improvement, the two variables that determine blast radius containment.
"The MCP server does not change what Zero Trust requires. It changes how long it takes to do it, and how much of it gets done consistently, rather than deferred because the friction was too high."
The real shift isn't any single workflow
It's what happens to every workflow at once: Zero Trust implementation stops being experienced as platform work and starts being experienced as conversation.
| Previous interaction model | MCP interaction model |
|---|---|
| Architect designs Protect Surfaces in a document, then navigates the platform to implement them, two separate activities with inevitable drift between them. | Design and implementation happen in the same conversation. The AI creates what the architect describes, in real time. |
| Transaction flow mapping is a manual documentation exercise, often deferred, often incomplete. | Flow mapping is a query and a confirmation. The AI identifies unmapped surfaces and creates flows on instruction. |
| Measure catalogue search requires platform navigation, results applied one at a time to each Protect Surface. | Catalogue search is a natural language query. Results are contextual, assignment is immediate. |
| Case management requires context switching. The analyst leaves the investigation to document it. | Case management is in-thread: create, note, escalate, and close without leaving the investigation context. |
| Zero Trust maturity state requires manual export and analysis to assess. | A current-state query returns a structured summary across all surfaces, flows, and measure implementation status in a single operation. |
| Implementation consistency depends on individual discipline. Every architect follows the process differently. | Consistency is enforced by the tool. The same query produces the same result regardless of which team member runs it. |
"Zero Trust fails in the gap between architecture and implementation. The MCP server closes that gap, not by changing the architecture, but by making the implementation as fluid as the conversation that designed it."
What you get out of the box
Across Zero Trust and Case Management domains, the full AUXO operational lifecycle.
Claude Desktop extension, binary, or container, each operational in under ten minutes.
macOS ARM/Intel, Linux AMD64/ARM64, and Windows, plus a container for team deployments.
For full capability. Zero Trust and Case Management domains are independently configurable.
Required to connect AUXO to Claude Desktop. The extension handles everything.
Every Zero Trust implementation, eventually.
Four steps from zero to operational
- Request your API token(s). Contact ON2IT support to request a Zero Trust token, a Case Management token, or both. Tokens are domain-specific, request only what you need. The server auto-detects which domains to enable based on which tokens are present at runtime.
- Choose your deployment path. For individual use with Claude Desktop: download the
.mcpbextension and double-click it. For VS Code Copilot or other clients: download the binary for your platform and add four lines to your MCP client configuration. For shared team environments: pull the container image and run it in HTTP mode behind a reverse proxy. - Verify the connection. Ask the AI assistant to "list all current protect surfaces." If the MCP server is connected and the token is valid, AUXO returns the current state. If the environment is empty, the AI will say so, and offer to help you create the first Protect Surface.
- Start with a gap analysis. The most immediately useful first query: "List all Protect Surfaces. For each one, show whether transaction flows are mapped and whether security measures are assigned. Flag any that have neither." This single operation surfaces the implementation gaps that most teams carry silently, and the AI can begin closing them in the same conversation.
Source code, binaries, and the Claude Desktop extension are available at github.com/on2itsecurity/auxo-mcp. API tokens are available on request from ON2IT support.
FAQ
What is the Model Context Protocol (MCP)?
MCP is an open standard that lets an AI assistant call external tools, APIs, and services as native operations inside a conversation, instead of only reasoning about them from training data or a pasted document.
What can the AUXO MCP Server do?
It exposes 27 tools across two domains: Protect Surface Management (creating and auditing Protect Surfaces, mapping transaction flows, assigning measures) and Case Management (creating, escalating, and closing incident cases). Any MCP-compatible assistant can call them directly.
Is the AUXO MCP Server open source?
Yes. It is released under the MIT licence and available at github.com/on2itsecurity/auxo-mcp, including pre-built binaries, a container image, and a Claude Desktop extension.
Which AI assistants does it work with?
Any MCP-compatible client, including Claude Desktop and VS Code Copilot. The Claude Desktop extension is the fastest path to a working setup; other clients connect through the pre-built binary or the container image.
Do I need both API tokens?
No. Zero Trust and Case Management are independent domains, each behind its own token. Request only the token for the domain you need, and the corresponding tools activate automatically.
Rob Maas is Field CTO at ON2IT, part of the CTO Office under CTO Lieuwe Jan Koning. He works on product strategy and architecture for AUXO, MDR, and ZTaaS, shaping ON2IT's Zero Trust customer journey and AI feature roadmap. He is also a Palo Alto Networks Cyberforce Hero.