One MCP interceptor screening every tool call against 27 pre-mapped rules across six jurisdictions — deterministically, before execution, with an audit trail your DPO can hand a regulator.
# Before Aegis Guard — opaque, unaudited, non-compliant @mcp.tool() async def screen_candidates(payload: dict) -> dict: return hiring_pipeline.evaluate(payload) # ───────────────────────────────────────────────────────────────── # After — one import, one decorator. Full EU AI Act Article 6 coverage. from aegis_guard import compliance_gate @mcp.tool() @compliance_gate # ← intercepts, classifies, blocks, and signs async def screen_candidates(payload: dict) -> dict: return hiring_pipeline.evaluate(payload) # Every call now produces a signed AuditRecord: # { # "risk_tier": "HIGH_RISK", "article": "Article 6(2) + Annex III §4", # "iso_controls": ["A.6.1.2", "A.9.2"], "is_blocked": true, # "hmac_sha256": "a3f8…c91d", "confidence": 0.95 # }
Aegis Guard doesn't just cover the EU AI Act. Every jurisdiction below is built on real, binding law — verified against primary-source legal text, not policy drafts or non-binding frameworks — and the rule set keeps growing as new jurisdictions are researched and activated. Five are already enforceable today; Colorado's is wired ahead of its 1 January 2027 effective date, flagged as a forward-looking risk rather than a live violation until then.
The EU AI Act came into force in August 2024, the UK's new automated-decision regime since 5 February 2026, and Nigeria, Singapore, and Egypt already enforce their own data-protection duties on automated processing. Every automated decision touching employment, credit, healthcare access, marketing, or biometric data is now a regulated event in at least one jurisdiction you operate in — whether your legal team knows the tool exists or not.
Subliminal manipulation, social scoring, real-time biometric surveillance in public spaces — Article 5 prohibited uses are absolute bans with fines up to €35M or 7% of global turnover. Without interception at the point of execution, your agents may already be crossing these lines.
CV screening, credit scoring, student assessment, law enforcement analytics — eight Annex III categories require full documentation of AI decision logic. A log file is not an audit trail. A JSONL record without cryptographic integrity proof is inadmissible evidence of compliance.
The EU AI Office is building tool-call-level inspection capability. MCP and agent frameworks are specifically in scope. "We didn't know the tool existed" is not a viable defence when your model context protocol server is actively running high-risk classification pipelines.
Enterprise certification programmes increasingly require ISO/IEC 42001 conformance. Annex A mandates explicit control mappings for AI risk (A.6.1), governance (A.9.2), and operational monitoring (A.10.1). Manual mapping exercises become instantly stale when your agent fleet evolves weekly.
Rather than auditing outputs after the fact, Aegis Guard intercepts every JSON-RPC tool call at the point of execution — before your tool's own business logic runs — classifying the intent, blocking prohibited operations, and writing a cryptographically signed audit record in a single atomic operation. Your compliance posture is enforced, not reported.
Compliance tooling usually forces a trade-off between what Legal needs, what Security will approve, and what Engineering will actually ship. Aegis Guard is designed so none of the three has to compromise.
Every evaluation — block, flag, or clean pass — produces a signed, timestamped record mapped to the exact statute and control that triggered it.
Aegis Guard runs in-process, inside your own perimeter — no payload ever leaves your environment to reach a third-party classifier.
Aegis Guard instruments an existing FastMCP tool without touching its business logic — you add an import and a decorator, nothing else moves.
@compliance_gate wraps any @mcp.tool() function directly, in place.Hover to see the stack separate — each layer below runs independently, in sequence, on every tool call.
Aegis Guard inserts itself at the MCP tool-invocation boundary using a single @compliance_gate decorator. Every tools/call JSON-RPC request is intercepted before Python execution begins. Tool name, all key paths, all nested values, and semantic conjunctions are extracted as structured signal vectors.
Existing agent implementations require no modification to business logic. One import and one decorator instruments an entire MCP server in under five minutes. The interceptor is deployed in your environment — no traffic leaves your perimeter.
Claude Desktop / OpenAI / custom agent sends JSON-RPC request to your MCP server.
{"method": "tools/call", "params": {...}}Payload is extracted before the decorated tool function is invoked.
_extract_all_text(payload) → signalsAll wired rules evaluated in-process. PROHIBITED tier triggers an immediate block.
_run_all_rules(tool_name, payload)HMAC-SHA256 record appended to append-only JSONL log. Optionally forwarded to your SIEM via webhook.
AuditRecord.from_compliance_result()Unlike probabilistic LLM-based classifiers, Aegis Guard uses a fully deterministic, explainable scoring model with four independent signal dimensions. Each dimension has a fixed maximum contribution weight, making every classification decision independently auditable and reproducible from the raw payload alone.
The engine evaluates every wired rule simultaneously — the 21 EU AI Act rules, Nigeria's NDPA 2023 rules, the UK's Data (Use and Access) Act 2025 rule, Singapore's PDPA 2012 rule, Egypt's PDPL rule, and Colorado's SB26-189 rule (wired now for forward-looking flagging; the underlying law itself isn't enforceable until 1 Jan 2027) — through the same scoring model. Every match returns a ScoreBreakdown object logging each dimension's contribution, the pre-clamp total, and whether an exception pattern reduced the score.
Every compliance evaluation — whether it results in a block, a flag, or a clean pass — generates an AuditRecord sealed with HMAC-SHA256 using a 32-byte minimum secret. The hash covers the canonical serialisation of the full record, making any post-hoc tampering immediately detectable. Each signed record can chain to the one before it in the same workflow — turn each link below to see the idea.
AI risk assessment — mandatory risk evaluation for high-impact AI operations
Human oversight — controls for human review of AI-generated decisions
Operational monitoring — continuous surveillance of AI system behaviour
Every enterprise has a different risk tolerance. A financial institution subject to FCA scrutiny may need stricter detection than a research lab operating under academic exemptions. Aegis Guard's operator threshold system lets your CISO dial sensitivity per EU AI Act tier without touching a single rule definition.
Three environment variables override the engine's per-rule confidence floor at runtime. Production environments enforce a safety invariant: thresholds cannot be inverted, and observe-only mode for PROHIBITED operations is hard-blocked.
Every rule shipped with Aegis Guard is pre-mapped to its governing Article or statute, risk tier, and ISO 42001 Annex A controls. Your compliance team does not write detection logic. They configure thresholds and review evidence.
| Tier | Rules | Jurisdictions | Example |
|---|---|---|---|
| Prohibited | 8 | EU | Real-time remote biometric identification in public spaces (Art. 5(1)(h)) |
| High Risk | 17 | EU · UK · Nigeria · Singapore · Egypt · Colorado | Employment, HR & recruitment scoring (Annex III §4); cross-jurisdiction automated-decision and consent-withdrawal duties |
| Transparency | 2 | EU | Chatbot disclosure and deepfake / synthetic media labelling (Art. 50) |
Nigeria's Data Protection Act 2023 governs cross-border transfers and processing after a data subject's objection — real, binding law, already enforced.
The UK repealed UK GDPR Article 22 and replaced it with a new automated-decision-making regime (Articles 22A–22D) under its own Act.
Singapore's Personal Data Protection Act requires an organisation to cease processing once an individual withdraws consent, unless another lawful basis applies.
Egypt's Personal Data Protection Law prohibits direct electronic marketing unless the sender provides a clear mechanism to refuse or withdraw consent. Its grace period ends around October/November 2026.
Colorado's Automated Decision-Making Technology Act requires deployers to give notice before use, explain adverse decisions, and offer meaningful human review for consequential decisions in employment, housing, lending, insurance, healthcare, education, and government services. Wired now so it flags as a forward-looking risk, not a live violation — the underlying obligations do not take effect until 1 January 2027.
A scheduled Regulatory Watch Loop checks each jurisdiction's primary regulator directly — the European Commission, GOV.UK / DSIT, Ireland's Data Protection Commission, Nigeria's NDPC & NITDA, Singapore's PDPC, and Egypt's Ministry of Communications and Information Technology — surfacing new developments for review before they become a gap in your compliance posture. This is regulatory intelligence: the early-warning layer that feeds rule development, separate from the rule engine itself. Enforced rule coverage today comes from five reviewed, activated rulesets — the EU AI Act, Nigeria's NDPA, UK GDPR under the Data (Use and Access) Act 2025, Singapore's PDPA, and Egypt's PDPL (26 rules, above); each new jurisdiction is added once its rule is built and reviewed. Ireland's coverage comes through the EU AI Act as an EU member state, with its own Data Protection Commission adding a dedicated monitoring layer on top. Colorado's rule is wired for forward-looking flagging ahead of its 1 January 2027 effective date (see above); Watch Loop coverage for its regulator is a planned future addition.
Your procurement cycle, your legal review cycle, and your engineering integration cycle all need to complete before the relevant enforcement date, whichever jurisdiction it falls in. Aegis Guard is designed to compress the integration timeline to days, not quarters.
An Aegis Guard Architecture Review is a structured 90-minute session with our compliance engineering team. We map your current MCP server topology against the jurisdictions you actually operate in, identify the highest-priority integration points, and deliver a written risk assessment and deployment plan your General Counsel can sign off on.
Not ready to talk? Try the live demo first ↗
Available for on-premise, VPC, and air-gapped deployments. NDA-protected review. · Privacy Policy · Security Overview