Technical writing on operating AI agents safely: access mapping, least-privilege drift, auditing Claude Code and MCP, and self-hosted governance.https://olivares.aienhttps://olivares.ai/blog/audit-evidence-offline-verificationhttps://olivares.ai/blog/audit-evidence-offline-verificationHow a hash-chained ledger with per-event Ed25519 signatures and OSCAL export produces audit evidence an external verifier can confirm independently.Sat, 11 Jul 2026 00:00:00 GMTauditledgerOSCALEd25519compliancehttps://olivares.ai/blog/governing-mcp-servers-rfc-9728-8707https://olivares.ai/blog/governing-mcp-servers-rfc-9728-8707Trace the OAuth 2.1 flow MCP mandates after a protected server returns 401, what RFC 9728 and RFC 8707 provide, and the security pitfalls.Sat, 11 Jul 2026 00:00:00 GMTMCPRFC 9728OAuthsecurityhttps://olivares.ai/blog/ephemeral-credentials-claude-code-workload-identityhttps://olivares.ai/blog/ephemeral-credentials-claude-code-workload-identityStatic API keys never expire, share identity, and silently shadow federation. WIF replaces them with attested JWT exchange for short-lived per-session tokens.Sat, 11 Jul 2026 00:00:00 GMTWIFephemeral-credentialsClaude Codeworkload-identityhttps://olivares.ai/blog/hooks-pep-deny-closed-policy-claude-codehttps://olivares.ai/blog/hooks-pep-deny-closed-policy-claude-codeHow a PEP classifies every Claude Code hook event into gating, context, or observe, wires the correct output schema per event, and deny-closes unknown events.Sat, 11 Jul 2026 00:00:00 GMTClaude Codehookspolicy-as-codedeny-closedhttps://olivares.ai/blog/permitted-vs-observed-drift-findinghttps://olivares.ai/blog/permitted-vs-observed-drift-findingDrift between what an AI agent is permitted to do and what it is observed doing becomes a structured, classified finding with confidence levels.Sat, 11 Jul 2026 00:00:00 GMTleast-privilegedriftaccess-mapobservabilityhttps://olivares.ai/blog/self-hosted-claude-apps-gateway-co-deploymenthttps://olivares.ai/blog/self-hosted-claude-apps-gateway-co-deploymentAnthropic's apps gateway handles OIDC auth for Claude. Your estate is more than Claude. This is the co-deployment architecture that governs both paths.Sat, 11 Jul 2026 00:00:00 GMTself-hostedClaudeco-deploymentarchitecturehttps://olivares.ai/blog/map-what-your-ai-agents-can-reachhttps://olivares.ai/blog/map-what-your-ai-agents-can-reachAn AI agent holds credentials, is permitted some actions, and is observed doing others. The access map shows the gap — and where least-privilege quietly broke.Fri, 05 Jun 2026 00:00:00 GMTagent-accessleast-privilegeobservabilityaudithttps://olivares.ai/blog/auditing-claude-code-and-mcp-self-hostedhttps://olivares.ai/blog/auditing-claude-code-and-mcp-self-hostedBuild an auditor-grade trail for Claude Code and MCP servers without leaving your perimeter: per-agent identity, a hash-chained ledger, untrusted MCP signals.Thu, 28 May 2026 00:00:00 GMTMCPClaude Codeauditself-hostedhttps://olivares.ai/blog/least-privilege-drift-for-ai-agentshttps://olivares.ai/blog/least-privilege-drift-for-ai-agentsHow to spot least-privilege drift in AI agents by comparing permitted access with observed behavior and enforcing policy at access time.Wed, 20 May 2026 00:00:00 GMTleast-privilegeai-agentspolicy-as-codeaudithttps://olivares.ai/blog/self-hosted-ai-governance-data-residency-gdprhttps://olivares.ai/blog/self-hosted-ai-governance-data-residency-gdprWhy self-hosting an AI platform is the strongest data-residency posture under GDPR: the governance tool never receives your data. Edges, not payloads.Wed, 13 May 2026 00:00:00 GMTdata-residencygdprself-hostedai-governancehttps://olivares.ai/blog/passive-discovery-vs-proxies-ai-agentshttps://olivares.ai/blog/passive-discovery-vs-proxies-ai-agentsTwo ways to see your AI agents: an inline proxy with high blast radius, or passive discovery from logs, OpenTelemetry and an eBPF backstop. An honest tradeoff.Wed, 06 May 2026 00:00:00 GMTpassive-discoveryobservabilityeBPFleast-privilege