Archive note: This post predates Mantle vocabulary canonicalization on 2026-05-15. Terminology here may use earlier language for install-time substrate-binding concepts. Current canonical vocabulary lives at Mantle Phase 1.

Sanctuary for Claude Managed Agents: Quickstart

You’ve deployed a Managed Agent. Now secure it with cryptographic identity, audit trails, and policy enforcement. Sanctuary Framework v0.7.0 added a smaller tool surface when this post was published; current releases expose 80+ tools.

Add Sanctuary in 30 Seconds

In your Managed Agent YAML config, declare the Sanctuary MCP server:

agent:
  model: claude-sonnet-4-5-20250514
  system_prompt: |
    You are a secure agent powered by Sanctuary Framework.
    Use the shr_generate tool to verify your identity status.
  tools:
    - type: agent_toolset  # Replace with your current agent toolset version
  mcp_servers:
    - name: sanctuary
      url: "npx @sanctuary-framework/mcp-server@0.7.0"

Deploy the agent. Sanctuary tools load automatically via MCP.

What Sanctuary Gives You

Sovereign Identity: Ed25519 cryptographic keypair + W3C DID per agent instance
Encrypted Audit Trail: Tamper-evident hash-chained logs; production audit checkpoints are currently unsigned until IC-05 closes; selective disclosure via zero-knowledge attestations
Principal Policy: Tiered role-based access control (RBAC) with time-locked capabilities
SIEM Export: Push audit events to any SIEM (Datadog, Splunk, Chronicle)
Sovereignty Health Reports: Current dashboard of identity status, policy compliance, audit integrity
Reputation Publishing: Cryptographic proof of track record via Verascore

Before vs. After

Aspect	Without Sanctuary	With Sanctuary
Identity	Implicit (cloud-managed)	Explicit (agent-owned Ed25519 keypair)
Audit	Default logging	Encrypted, tamper-evident hash-chained logs; production audit checkpoints are currently unsigned until IC-05 closes; SIEM-ready
Policy	Platform-level (static)	Principal-level (dynamic, time-locked)
Disclosure	All-or-nothing	Selective (ZK attestations)
Reputation	Opaque	Verifiable via Verascore

Verify Sanctuary is Ready

Call the manifest tool from your agent logic. If you see the Sanctuary tool surface listed, Sanctuary is active. Current releases expose 80+ tools:

# Verify Sanctuary tools are available
manifest = agent.call_tool("manifest")
print(f"Sanctuary loaded: {len(manifest['tools'])} tools available")

# Key tools to verify:
# - identity_create: Generate sovereign identity (Ed25519 keypair + DID)
# - shr_generate: Generate Sovereignty Health Report
# - audit_export_siem: Export audit logs (CEF/OCSF)
# - reputation_publish: Publish to Verascore

Note: Exact calling syntax depends on your agent framework. The above is illustrative.

Or check logs: you’ll see MCP server initialization and tool binding on startup.

Next Steps

Identity-Gate Your Agent: Call identity_create on startup. Store the keypair securely.
Log to SIEM: Configure audit_export_siem with your SIEM endpoint (Datadog, Splunk, etc.)
Publish Reputation: Use reputation_publish after key milestones (e.g., 1000 verified transactions).
Add Negotiation (optional): Install pip install concordia-protocol to enable multi-agent policy negotiation.
Monitor Compliance: Call shr_generate in your observability dashboard.

Resources

Sanctuary Framework: https://github.com/eriknewton/sanctuary-framework
Verascore Reputation: https://verascore.ai
Managed Agents: https://docs.anthropic.com/en/docs/build/agents
MCP Protocol: https://modelcontextprotocol.io

Launched April 8, 2026. Sanctuary Framework is open source. Build agents that own their identity and prove their integrity.