Blekline operates before model and tool execution. These partners operate after or alongside for quality and output validation.
Positioning
| Product | Layer | Role |
|---|---|---|
| Blekline | L4 | Mask, enforce, audit at MCP ingress |
| LangSmith | L3 | Eval, traces, datasets — consumes audit metadata |
| Guardrails AI | L3/L4 | Post-model output validation — optional second layer |
Data flow
Agent → Blekline (allow/block) → Model/Tools → Guardrails (optional output check) → User
↓
LangSmith (eval traces, metadata only)
LangSmith is not a SIEM. For compliance export see Enterprise telemetry.
Next steps: Trust boundaries · Telemetry spec · AI Enablement Stack · Open workspace