Gives Claude the ability to detect infrastructure anomalies, trigger auto-remediation workflows, and roll back failed deployments. The server surfaces health check primitives and self-healing hooks that you'd wire into your monitoring stack when you want LLM-driven incident response instead of manual runbooks. Part of MEOK AI Labs' suite, with built-in EU AI Act compliance scaffolding for risk management and transparency documentation. You'd reach for this if you're building autonomous ops pipelines where Claude needs to assess system health, decide whether to remediate or escalate, and execute rollback procedures without human intervention.
The only self-healing infrastructure MCP with cryptographic governance.
Every remediation is Ed25519-signed, BFT Council-approved, and EU AI Act compliant.
| Feature | What It Means |
|---|---|
| Sigil Bus (Ed25519) | Every remediation action is cryptographically signed. Tamper-proof audit trail. |
| BFT Council | Critical decisions require 3-of-5 Byzantine Fault Tolerant consensus. No single point of failure. |
| 28-Domain Compliance | Auto-remediation rules vary by industry domain (Finance, Healthcare, Energy, etc.). |
| EU AI Act Built-In | Articles 9, 13, 15, 26, 50 — risk management, transparency, bias, FRIA, watermarking. |
| NIST AI RMF | Govern, Map, Measure, Manage — all four functions implemented. |
| ISO 42001 | AI management system controls mapped to auto-remediation policies. |
# Install via pip
pip install sovereign-self-healing-mcp
# Or via Smithery
npx -y @smithery/cli@latest install sovereign-self-healing-mcp --client claude
# Or via uvx (fastest)
uvx sovereign-self-healing-mcp
| Tool | Description | Governance |
|---|---|---|
node_health_check | Check CPU, memory, disk, GPU | Auto + attested |
cluster_health_check | Check all 9 nodes | Auto + attested |
restart_service | Restart failed service | Auto if safe, council if critical |
auto_remediate | Fix common issues | 28-domain compliance check first |
gpu_orchestration | Allocate GPU across cluster | Auto + attested |
failover_decision | Determine failover action | BFT Council vote |
cost_report | Infrastructure cost breakdown | Read-only |
compliance_check | Check if issue is auto-remediable | 28-domain rules |
council_propose | Propose critical action to BFT Council | Governance layer |
council_vote | Cast vote on pending proposal | Governance layer |
council_status | Check proposal status | Governance layer |
verify_attestation | Verify Ed25519 sigil on any action | Cryptographic proof |
Every tool call goes through the Sovereign Governance Layer:
Issue Detected → Domain Classification → Compliance Check → [Auto | Council | Block] → Sigil Sign → Log
disk_full → Auto-remediate. data_loss → BFT Council.backup_verify → Auto. phi_exposure → BFT Council + manual review.load_balancer_shift → Auto. grid_disconnect → BFT Council + safety isolation.This MCP server is part of CSOAI Sovereign Temple — the open-source sovereign AI governance infrastructure.
| Component | Purpose |
|---|---|
| Sigil Bus | Ed25519 attestation chain for every agent action |
| BFT Council | Byzantine Fault Tolerant consensus for governance |
| 28-Domain Engine | 280 data sources, 28 industry domains, $0 cost |
| Horus Intel | Weekly competitive intelligence for AI governance |
| HIVE 1 | Enterprise compliance attestation (51 MCP servers) |
MIT © CSOAI-ORG · UK Companies House 16939677
Built with 💜 by CSOAI Labs · Sovereign AI Governance
silenceper/mcp-k8s
azure/containerization-assist
io.github.evozim/aws-builder
reza-gholizade/k8s-mcp-server
flux159/mcp-server-kubernetes