
Pre-Inference MCP & Agent Firewall
Inspects every Model Context Protocol (MCP) tool invocation, enforcing per-environment allowlists and blocking indirect prompt-injection payloads before execution.
Audits cadrés, revues de sécurité IA et surveillance gérée réalisés par nos ingénieurs sécurité. (Architecture review of LLM applications, agents and MCP integrations.)

Inventory every agent, model, owner and MCP server in one place.
Define exactly which tools each agent may call, per environment.
Detect indirect injection in user input and retrieved content.
API keys, tokens and personal data are removed before inference.
Writes, payments and shell commands are routed for step-up approval.
Every allow, block and approval is recorded with its rationale.

Inspects every Model Context Protocol (MCP) tool invocation, enforcing per-environment allowlists and blocking indirect prompt-injection payloads before execution.

Automatically redacts bearer tokens, API keys and personal data before model inference and writes every allow/block decision to an append-only ledger.
CSH-RULE-001..004) mapped to MITRE ATT&CK (T1110.001, T1098.003, T1048, T1190) with idempotent dedupe_key ingestion and SHA-256 immutable event hashes.[REDACTED_SECRET], [REDACTED_PII]), explicit per-agent Model Context Protocol (MCP) tool allowlists, and indirect prompt-injection blocking.ENABLE / FORCE ROW LEVEL SECURITY across all 20 tenant tables bound to transaction-local current_tenant_id(); cross-tenant queries fail closed with HTTP 403.isolate_host) and credential revocation require named human step-up approval recorded in an append-only audit trail.DNS TXT or HTTP ownership token required before scanning. Continuous evidence aligned with ANSSI Guides d'hygiène informatique, Directive NIS2 France, RGPD / CNIL.ML-KEM / ML-DSA), and 2035 full PQC completion.FORCE ROW LEVEL SECURITY on all 20 tenant tables.ALTER TABLE security_events ENABLE ROW LEVEL SECURITY; ALTER TABLE security_events FORCE ROW LEVEL SECURITY; CREATE POLICY tenant_isolation_policy ON security_events FOR ALL TO csh_app_role USING (organization_id = current_tenant_id()) WITH CHECK (organization_id = current_tenant_id());
Non. La détection repose sur des règles déterministes. ARYA SAI produit uniquement des synthèses consultatives à partir de preuves expurgées ; chaque action exige une validation humaine.
Vous choisissez l'UE (Francfort) ou le Royaume-Uni (Londres) lors de l'intégration.
Oui, CSH Compliance relie automatiquement les preuves techniques aux contrôles ISO/IEC 27001:2022, NIS2 et RGPD.
Uniquement les actifs dont vous avez prouvé la propriété par jeton DNS TXT ou HTTP.
Oui, avec gestion déléguée des locataires et licences mutualisées.

30 minutes avec un ingénieur sécurité. Sans slides.