PRE_GA · PILOT_OPEN
>_ PROTECTEDCODE v1.0  ·  target:.NET/C#/VS2026  ·  boundary:azure_landing_zone  ·  egress:none_by_default  ·  [WORKING_DESIGN]

AI-ASSISTED DEVELOPMENT, inside your Protected boundary.

PROTECTEDCODE runs the coding model, retrieval index, and build/test tools inside your department's own Azure Landing Zone. Source code, prompts, embeddings and retrieved context never reach a public AI service.

PROTECTED A/B CLIENT-OWNED AZURE LZ VISUAL STUDIO 2026 BYOM NO PUBLIC EGRESS OFFLINE LICENSING
PROTECTED_A > PROTECTED_B > AZURE_LANDING_ZONE > VISUAL_STUDIO_2026 > ENTRA_ID > RAG_INDEX > BUILD_TEST_SANDBOX > SIGNED_AUDIT_LOG > OFFLINE_LICENSING > PROTECTED_A > PROTECTED_B > AZURE_LANDING_ZONE > VISUAL_STUDIO_2026 > ENTRA_ID > RAG_INDEX > BUILD_TEST_SANDBOX > SIGNED_AUDIT_LOG > OFFLINE_LICENSING

// 01

THE PROBLEM WITH PUBLIC AI ON PROTECTED CODE

$ diff public-ai protectcode
STATUS_QUO

Public / SaaS AI Coding Tools

unapproved for Protected classification
  • Source code and prompts leave the department's boundary
  • Classification review blocks adoption for Protected A/B work
  • No department control over model, logging, or retention
  • Per-seat SaaS billing, token overage risk
  • No reusable SA&A evidence — every review starts cold
PROTECTEDCODE

Protected AI Coding Platform

deployed inside your Azure Landing Zone
  • Coding model, RAG index and build/test run in your Azure LZ
  • No public AI service in the Protected data plane
  • Department controls identity, network, logging, keys
  • Annual platform licence, not per-token billing
  • Reusable SA&A evidence package per deployment variant

"You must consider the security classification of the code." — Treasury Board Secretariat, Guide on the use of generative AI


// 02

REQUEST FLOW · END-TO-END

$ trace request --verbose
DIAGRAM_01 Protected AI Coding Data Flowhandle: arch/01v1.0
VISUAL STUDIO2026 · BYOM provider
[1] private HTTPS
AI GATEWAYprivate endpoint · Entra auth
[2] policy_check
MODEL ROUTINGfast model / reasoning model
[3] retrieve
RAG / CODE INDEXproject-scoped · no cross-project
[1] REQUEST
[2] AUTH
[3] POLICY
[4] RETRIEVE
[5] SANDBOX_EXEC
[6] AUDIT
[7] RESPOND
No public AI dependency in the data plane. Departmental workstation-to-Azure connectivity (ExpressRoute, VPN/GC network, hub firewall, DNS) is the department/SSC's existing responsibility — the product boundary begins at the AI Gateway's private endpoint. Build/test agent actions run in disposable, no-Internet sandboxed workers with tool allowlists and diff preview before write.
$ example --provenance
AI ANSWER
CustomerStatus appears to be calculated in CustomerEligibilityService.cs.

Sources used:
  CustomerEligibilityService.cs
  CustomerRepository.cs
  EligibilityRules.cs
  Microsoft EF Core docs (approved local copy)
  Department Application Architecture Guide

Model: <approved model/version>
RAG index: <project/version>
Public AI service used by platform: NO

// 03

CAPABILITY NODES

6 of 6 active
NODE_01● ACTIVE

Private Inference & Embeddings

Coding and reasoning models run on approved compute inside your boundary. Embedding generation is local — code chunks never hit an external embeddings API.

egress = none
NODE_02● ACTIVE

Project-Isolated RAG

Per-project/repository indexes by default. The gateway maps identity and current solution to allowed index IDs — no global cross-project vector store.

index.scope = project
NODE_03● ACTIVE

Sandboxed Agent Execution

Build/test/agent tools run in disposable, no-Internet workers with least-privilege identities, tool allowlists, and diff preview before any write.

worker.egress = none
NODE_04● ACTIVE

Prompt-Injection Defense

Retrieved repository content — READMEs, comments, package metadata — is always untrusted. System/tool policy outranks anything retrieved from source.

retrieval.privilege = none
NODE_05● ACTIVE

Provenance & Audit

Every answer cites the files and docs it drew from, plus model/version and RAG index. Tool calls, exit codes, and diffs are logged to your SIEM.

audit.sink = departmental_siem
NODE_06● ACTIVE

Offline Licence Validation

Signed licence file, validated locally against an embedded public key, bound to tenant/subscription identifiers. No supplier call-home dependency.

license.callhome = false

DEPARTMENT / SSC PROVIDES

  • Managed developer workstation & device policy
  • GC/department network connectivity to Azure
  • ExpressRoute / VPN / GC network transport
  • Hub routing, firewall, approved source networks
  • Private DNS forwarding / enterprise name resolution
  • Landing-zone subscription / VNet provisioning

PROTECTEDCODE PROVIDES

  • Private AI API / Gateway endpoint
  • Entra-aware authentication and project policy
  • Model / backend routing
  • Project-isolated RAG / code intelligence
  • Build/test agent controls and sandboxing
  • Platform audit/provenance and sanitized diagnostics

// 04

DEPLOYMENT PROFILES BY CLASSIFICATION

$ ls /profiles
[PROTECTED_A]

Private workload, no public ingress, private AI processing. Management-plane allowlist may be permitted; data plane can remain no-egress. Azure landing zone is a natural fit once authorized.

egress: management-plane allowlist only
[PROTECTED_B]

Stronger segmentation, private endpoints, centralized logging, deny-by-default. Explicit client-approved exceptions only — no direct data-plane Internet path.

egress: deny-by-default / explicit exception
[PROTECTED_C+]

Special high-security architecture. Treated as a later/special deployment target, not the initial Azure product profile — assume no egress until specifically authorized.

egress: none until authorized

// 05

INTEGRATION TARGETS

$ which ide
● PRIMARY
01

Visual Studio 2026

Bring Your Own Model — custom OpenAI-compatible/Ollama-compatible endpoint. Chat, agent edits, and inline completion against your private endpoint only.

provider: Ollama-compatible
● TARGET
02

Azure Landing Zone

Deploys as a workload/spoke into your existing department-owned or SSC GC Cloud One-managed landing zone. Your subscription, your identity, your logging.

deploy: IaC / Bicep
● TARGET
03

Approved Knowledge Repository

.NET/C#, ASP.NET Core, EF Core, SQL Server docs, approved Microsoft Learn material, and your own department coding standards — imported, chunked, retrieved.

mode: RAG, not retraining
● TARGET
04

Departmental SIEM

Signed, append-only audit of accepted requests, tool calls, and policy decisions — routed to your existing centralized logging, not a supplier-held log.

sink: existing SIEM
● TARGET
05

Entra ID

Authentication and project-scoped authorization run through your existing Entra tenant. No separate identity system for developers to manage.

auth: Entra-aware
● PLANNED
06

Java / IntelliJ

Initial technical focus is .NET/C#. Java and IntelliJ integration follow the same gateway/RAG pattern once the .NET profile is validated.

status: roadmap

// 06

SECURITY POSTURE

no direct internet egress · deny by default · least privilege
01

Private inference & endpoints

Azure services reached through private endpoints where feasible; public network access disabled where supported.

02

No public IPs

AI compute, RAG, build/test and storage do not require public IP addresses.

03

Deny by default

Routes, NSGs, and firewall rules follow deny-by-default / allow-by-exception.

04

Controlled logging

Logs stay in approved departmental SIEM; prompt-body logging minimized/disabled unless explicitly required.

05

Separated management plane

Signed model packages, docs, and NuGet updates use a controlled import route — distinct from prompt/code traffic.

06

Secret exfiltration controls

No direct network from build/test workers by default; secret scanning/redaction; destination allowlist.

THREATEXAMPLECONTROL
Repository prompt injectionREADME/comment tells agent to ignore policy or execute a commandRetrieved content is always untrusted — never becomes a privileged instruction
Malicious dependency contentVendored package targets the agent with embedded instructionsApproved package provenance; sandboxed tool execution
Cross-project leakageQuery retrieves protected chunks from another projectPer-project indexes; no global cross-project vector store by default
Destructive actionAgent deletes files or runs an unexpected commandTool allowlist, diff preview, confirmation gates, rollback

// 07

LICENSE TIERS

$ cat /etc/protectcode/license

Optimized for Government of Canada procurement. Annual platform/environment licence, included developer seats, additional seats priced per developer. You pay your own Azure compute/storage — no token billing.

TIER_01

PILOT

single environment · pilot · sole-sourceable
$12,000 /fy
+ HST · well under GC $25K threshold
  • 1 Azure Landing Zone deployment
  • Up to 10 named developers
  • .NET/C# + Visual Studio 2026 BYOM
  • Approved Knowledge Repository (RAG)
  • Standard signed audit log
  • Email support · 1 contact · 24hr response
  • Single department · pilot environment
[START_PILOT]
TIER_03

ENTERPRISE

multi-department · regulated
$95,000 /fy
competitive process · IBD set-aside eligible
  • Unlimited environment deployments
  • Unlimited named developers
  • HA/DR reference architecture
  • Java/IntelliJ roadmap priority
  • Dedicated architect · onboarding + quarterly review
  • 24/7 support · 4 contacts · 1hr Sev-1
  • Multi-department · enterprise-wide
[CONTACT_SALES]
add-ons · separate engagement · SOW-based
  • Landing-zone deployment & SA&A evidence package · fixed-fee per environment
  • Custom Approved Knowledge Repository authoring · day rate
  • Model evaluation / benchmarking harness setup · fixed-fee
  • Multi-year option terms · base year + option years
supported_config · Azure Landing Zone (department or GC Cloud One) · Windows/Linux compute · Visual Studio 2026+
support_covers · PROTECTEDCODE platform defects + documented configuration · within named-contact limits per tier
out_of_scope · departmental network/transport · endpoint management · GPU/compute cost (customer-paid) · contract authoring (available as PS)
customer pays own Azure GPU/compute/storage · fiscal-year aligned billing (Apr 1–Mar 31) · pro-rated first term available

// 08

OPEN A SECURE CHANNEL

tls 1.3 · ed25519
$ ssh demo@protectedcode.ca

The authenticity of host 'protectedcode.ca' can't be established.
ED25519 key fingerprint is SHA256:pr0t3ct/c0de//v1.0
Are you sure you want to continue? [yes]

> Initiate engagement.

Architecture reviews, technical demos, and a proof-of-concept plan scoped to a 20–25 developer pilot. Tell us your landing zone and classification target and we'll walk through the deployment demarcation.

REQUEST_PILOT_BRIEFING  handle: form/demo