PROXY:AI

1. A knowledge-accounting layer for monitoring and budgeting the use of AI resources
2. Operational event registration, administrator analytics dashboards and provider-error monitoring
3. Defining rules that allow AI actions, require approval or block tools
4. A confidentiality mode for working with sensitive data
5. An MCP gateway with multi-level tool-access control at user, role and team level
6. Restricting and auditing access to tools, functions, code execution and terminals
7. Configuring AI request and response transformations without code through simple administrative rules

1. Controlled AI use through audits, access policies, human oversight, decision logging and model-usage documentation
2. Readiness for EU AI Act requirements, including visibility of models, policy decisions, events, risks and audit trails
3. Managing personal-data exposure by detecting, masking, blocking or escalating requests
4. Registers and operational evidence useful for AI management under ISO 42001
5. Regulatory compliance templates for the EU AI Act
6. An SBOM model and compliance reporting for Cyber Resilience Act requirements
7. Human Oversight mechanisms including AI-content labeling, FRIA/ASIA export and approval gates for critical actions
8. A compliance panel presenting policies, access decisions, events, risks and active models

1. A unified management point for communication with LLMs by applications, agents and users
2. Support for multiple cloud, local and on-premises AI model providers
3. Mapping errors from different providers to one format compatible with the OpenAI API
4. Routing models according to configuration, availability, cost, policies and user permissions
5. Visualization of a request's route through rules, models, providers and policy decisions
6. Fallback-model configuration and automatic switching when the primary model is unavailable
7. A model catalog with metadata including alias, provider, type, status and usage mode
8. Per-tier SLA isolation with dedicated instances and hardware-level GPU isolation
9. Declarative regional data residency in an EU-only per-tenant model
10. Defining AI resources as Custom Resource Definitions in Kubernetes
11. Protection against excessive spending by detecting unusual consumption patterns and terminating suspicious sessions
12. A dry-run panel for testing ABAC security rules before production activation
13. No-code AI request and response transformations
14. IDE integration through a dedicated plugin with real-time cost and token visibility
15. An LLM Playground for experiments, response comparison and parameter tuning
16. A Python and WASM plugin framework extensible through a marketplace
17. Model validation before launch in the AI environment
18. An internal-currency model for accounting for AI use, including per-action costs, budgets and transfers

1. Model-access decisions based on user, group, model, data-type and environment attributes
2. Predefined security-policy profiles for regulated sectors including finance, healthcare, public administration and energy
3. Real-time detection of abuse attempts including injection, jailbreak, prohibited content and token-consumption anomalies
4. Measuring the cost of every request by user, team, project and model with hierarchical budgets
5. A central versioned prompt registry
6. Prompt-template support and request decoration with central gateway rules
7. Caching recurring requests using semantic similarity with per-tenant key isolation

1. A dashboard for database-integration status, outbox/reconciliation, cost and usage
2. A conversational Knowledge Chat assistant supporting operators or administrators
3. A model registry with test statuses, BYOK tests and routing projection
4. AI operation logging and auditing with filters and CSV/JSON export
5. Unified telemetry with replay of historical AI requests
6. Virtual-key management covering creation, synchronization, masking, rotation and auditing
7. ABAC rule administration with YAML import/export, dry-run mode and gateway publishing
8. An ABAC decision dashboard, HITL queue, user logs and risk register
9. AI security controls through guardrails, control policies, human review and evidence logs
10. Operational schedules, Langflow agents and runnable agent lists
11. Spending and token logs for cost reporting and configuration
12. Docker log browsing with a time filter
13. Orchestration of complex multi-step agent processes with state retained between steps
14. Automatic or semi-automatic model evaluation and adaptive real-time A/B prompt tests

2026/Q3IN PROGRESS

2026/Q4

It is the central control layer — one router for different AI models. It addresses shadow AI and protects the organisation against sensitive data leaks and uncontrolled costs

FinOps and Anti-DoW mechanisms set tenant and user limits, including protection against denial-of-wallet attacks

PROXY:AI supports human-in-the-loop oversight: a critical process pauses until the appropriate person authorises it

The immutable WORM audit log records the complete operational history: model, configuration, invocation cost and data needed to reconstruct the decision

The MCP Gateway lets agents use only explicitly assigned tools, such as Jira, and records an audit trail of their actions

Yes. Model responses are scanned again for sensitive data and dangerous instructions

Yes. The dry-run panel allows ABAC rules to be tested on simulated traffic before production deployment

Yes. The AI Firewall analyses data in under 50 milliseconds and neutralises items such as Polish national ID and tax numbers and email addresses before they leave the organisation's network

Semantic caching serves repeated questions with similar meanings from the cache, without querying an expensive model again

Real-time guardrails filter requests and block attacks such as prompt injection and jailbreaks

Zero-Code Switch: change the primary model in PROXY:AI settings without rewriting business application code

It does not admit requests blindly. For every request, it evaluates who is asking, from which environment, using which data and at what risk — then allows, restricts or blocks the operation