Three-Tier Framework
Public read-only Three-Tier Framework MCP for classifying systems across Database, Automation, and Judgment.
Transport http, sse
Tools 6
Eval status pass
Last review 2026-05-10
Risk Summary
Read-only advisory framework surface. It returns structured analysis and does not access private tenant state.
Evidence
The public card covers read-only framework classification, debugging, MCP analysis, policy artifact identification, metaphor mapping, and architecture comparison.
Install Snippets
Codex toml
[mcp_servers."three-tier-framework"]
url = "https://framework.mcp.createsomething.agency/mcp"Claude Desktop / Code json
{
"mcpServers": {
"three-tier-framework": {
"url": "https://framework.mcp.createsomething.agency/mcp"
}
}
}Claude Code CLI shell
claude mcp add --transport http three-tier-framework https://framework.mcp.createsomething.agency/mcpCursor json
{
"mcpServers": {
"three-tier-framework": {
"url": "https://framework.mcp.createsomething.agency/mcp"
}
}
}Eval Gate
- Suite:
braintrust:eval:mcp:public-trust - Status:
pass - Required checks: auth_not_required, endpoint_reachable, expected_tools_present, grounded_content, latency_budget, no_credential_material, tool_call_ok, tools_listed
Observability
- Braintrust: three-tier-framework-mcp / three_tier_framework_public_trust
- Langfuse: three-tier-framework-mcp / production
- langfuse: label_declared
Limitations
- Framework outputs are advisory and must be checked against the actual system being evaluated.
- Public evidence summarizes eval outcomes only; raw traces are not exposed.
- The server does not grant access to private client policies or tenant data.
Redacted Samples
config/public-trust/samples/mcp-three-tier-framework-classification.md- Framework classification rollup
Escalation: [email protected]
External Listings
No external mirror has been published for this card yet.