This matrix maps a representative enterprise AI-NOC RFP/RFI requirement set onto the section of this document that addresses each requirement, plus a compliance disposition consistent with standard vendor-response conventions:
- Full Comply — feature is in production today, source-verified, no caveats
- Intend to Comply — clear roadmap, partial implementation in v2.0.63, or deployment-dependent (e.g., SLA contracts)
- No Comply — not currently in scope; customer should plan accordingly
The requirement IDs (REQ_X.YYY) follow a typical AI-NOC RFP structure (Registry & Asset Management, Unified Gateway, Observability & Telemetry, Governance & Infrastructure, Security & Access Management, Sandbox & Developer Platform). Use this matrix as a starting point when responding to similar enterprise RFPs.
| REQ |
Disposition |
Evidence in this doc |
| REQ_1.001 — Support all major commercial LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure OpenAI, Cohere, Mistral, Together, Groq) via a single API |
Full Comply |
§4.1 — 14 providers preconfigured, all listed |
| REQ_1.002 — Route to self-hosted / open-source models (LLaMA, Mistral) on vLLM, Ollama, Triton, TGI |
Full Comply |
§4.1 (Custom + Ollama provider types — application code unchanged) |
| REQ_1.003 — Unified OpenAI-compatible HTTP API (base-URL swap) |
Full Comply |
§3.1, §37 |
| REQ_1.004 — Centralized model catalog with per-team provisioning + per-model aliases |
Full Comply |
§4.2, §15 (ModelAccess per team/tenant) |
| REQ |
Disposition |
Evidence |
| REQ_1.101 — Centralized catalog for agents w/ metadata, owner, collaborators, version, usage |
No Comply |
§43.1 — Agent Registry |
| REQ_1.102 — Publish shareable agents (bundle prompts + tools + params) |
No Comply |
§43.1 |
| REQ_1.103 — Fine-grained RBAC on Agent Registry |
No Comply |
(RBAC engine exists §17; no Agent Registry to scope it to) |
| REQ_1.104 — Agent app versioning with safe upgrades + rollback |
No Comply |
§43.1 |
| REQ_1.105 — Trace every agent request |
Full Comply |
§21 (mandatory OTel), §22 |
| REQ |
Disposition |
Evidence |
| REQ_1.201 — Connect to pre-built/custom MCP servers w/ metadata + auth |
Full Comply |
§14 |
| REQ_1.202 — Centralized registry w/ discovery + config + session mgmt + fine-grained authZ |
Full Comply |
§14 (ACLs gating discovery + invocation) |
| REQ_1.203 — Monitoring + observability (volume, latency, error rates, audit logs) |
Full Comply |
§14 (independent MCP observability), §21 |
| REQ_1.204 — Virtual MCP servers from curated tool subsets |
No Comply |
§43.1 — Virtual MCP Servers |
| REQ |
Disposition |
Evidence |
| REQ_1.301 — Versioned prompt/skill registry w/ Dev/Staging/Prod + audit trails |
No Comply |
§43.1 — Prompt Registry |
| REQ_1.302 — Continuous offline/online eval against test sets w/ scoring |
No Comply |
§43.1 — Prompt Evaluation Framework |
| REQ |
Disposition |
Evidence |
| REQ_1.401 — HITL approval workflow for Tier-3 tool calls |
No Comply |
§43.1 — Human-in-the-Loop Workflows |
| REQ |
Disposition |
Evidence |
| REQ_1.501 — Handle 50k / 100k / 200k token contexts without silent truncation |
Full Comply |
§3 (transparent passthrough); model registry tracks per-model context limits §4.2 |
| REQ_1.502 — Short-term + long-term memory mgmt |
Intend to Comply |
Session embeddings + vector store exist (LanceDB §34); first-party agent-memory abstraction is on roadmap |
| REQ |
Disposition |
Evidence |
| REQ_2.001 — Perimeter security (IP whitelist, rate limit, DDoS) |
Full Comply |
§24, §6, §35 |
| REQ_2.002 — WAF for OWASP Top 10 (SQLi, XSS, etc.) |
Full Comply |
§24 |
| REQ |
Disposition |
Evidence |
| REQ_2.101 — Model-agnostic abstraction |
Full Comply |
§3, §4 |
| REQ_2.102 — Built-in failover w/ latency-threshold-driven re-route |
Full Comply |
§8 (circuit breakers, health monitoring, priority fallback) |
| REQ_2.103 — QoS routing for latency-sensitive workloads |
Full Comply |
§5.1 (Priority/QoS routing + per-route SLA hints) |
| REQ_2.104 — Centralized budget enforcement |
Full Comply |
§6.2 (4-tier budget) |
| REQ |
Disposition |
Evidence |
| REQ_2.201 — Mandatory access path via whitelisted tool discovery |
Full Comply |
§14 (ACL-gated tools/list) |
| REQ_2.202 — RBAC-bound virtual servers |
Intend to Comply |
RBAC + per-server ACLs ship today; virtual-server construction is on roadmap (§43.1) |
| REQ_2.203 — Acryl DataHub lineage tracking |
No Comply |
§43.1 (Kafka stream allows external lineage reconstruction) |
| REQ |
Disposition |
Evidence |
| REQ_2.301 — Multi-agent orchestration + sandbox for dynamically generated code |
No Comply |
§43.1 — Multi-Agent Orchestration |
| REQ_2.302 — Agent profile mgmt + recursion-depth limits |
No Comply |
§43.1 |
| REQ |
Disposition |
Evidence |
| REQ_2.401 — Configurable rate limits across models/users/teams/keys/apps/tools/servers |
Full Comply |
§6.1 |
| REQ_2.402 — Fallback options + budget alerts |
Full Comply |
§6.1, §6.2 |
| REQ |
Disposition |
Evidence |
| REQ_2.501 — Embedding-similarity or exact-match cache |
Full Comply |
§7 |
| REQ |
Disposition |
Evidence |
| REQ_2.601 — Policy-based routing (QoS, cost, priority, latency) incl. semantic / weighted / round-robin / cost-optimized |
Full Comply |
§5.1 |
| REQ_2.602 — Prompt compression (token reduction) |
No Comply |
§43.1 — Prompt Compression |
| REQ |
Disposition |
Evidence |
| REQ_3.001 — Mandatory OTel traces for all tool calls |
Full Comply |
§21 |
| REQ_3.002 — Traffic classification + anomaly detection for control/orchestration/management traffic |
Intend to Comply |
Event streams ship today (§22); anomaly detection on those streams is the customer’s SIEM rule set |
| REQ_3.003 — Secure link for all management traffic (confidentiality + integrity) |
Full Comply |
§35 (HTTPS, mTLS, Redis TLS) |
| REQ_3.004 — No sensitive info in logs in cleartext |
Full Comply |
§35 (sensitive-log scrubbing via DLP on emit) |
| REQ |
Disposition |
Evidence |
| REQ_3.101 — Comprehensive logging of prompts, responses, metadata, tokens, cost, latency |
Full Comply |
§21, §22 |
| REQ_3.102 — End-to-end distributed tracing w/ OTel |
Full Comply |
§21 |
| REQ_3.103 — Real-time + historical metrics w/ configurable alert thresholds |
Full Comply |
§21 (Prometheus + Langfuse) |
| REQ_3.104 — Filter/aggregate by user/org/custom tags |
Full Comply |
§21, §22 (ClickHouse) |
| REQ_3.105 — RAG-specific metrics (Context Precision / Recall / Faithfulness) |
No Comply |
§43.1 — RAG Quality Metrics |
| REQ_3.106 — Context window usage visualization + “lost in the middle” detection |
Intend to Comply |
Token counts captured per request (§21); the analytical lens on context efficiency is a UI roadmap item |
| REQ_3.107 — Monitor MCP server uptime/error/latency independently of LLM |
Full Comply |
§14 |
| REQ |
Disposition |
Evidence |
| REQ_3.201 — End-to-end distributed tracing per request |
Full Comply |
§21 |
| REQ_3.202 — Structured JSON logs w/ trace_id, user_id, team_id, endpoint, status, latency_ms |
Full Comply |
§21 |
| REQ_3.203 — Separate log streams for auth failures + rate limit; only redacted prompts |
Full Comply |
§21, §22 (topic separation), §18 |
| REQ_3.204 — OTel SDK, Jaeger/Tempo, Datadog APM, Prometheus, ELK/Splunk |
Full Comply |
§21 |
| REQ |
Disposition |
Evidence |
| REQ_3.301 — Alert on refusal/safety-filter trigger drift vs. baseline |
No Comply |
§43.1 — Model Drift Detection |
| REQ |
Disposition |
Evidence |
| REQ_3.401 — Detect + block prompt injection at gateway (semantic guard, keyword/syntax patterns, anomalous behaviour) |
Full Comply |
§19 |
| REQ_3.402 — Strict output schema validation + reject/re-prompt |
Full Comply |
§27, §19 |
| REQ |
Disposition |
Evidence |
| REQ_3.501 — FOCUS-compliant export |
No Comply |
§43.1 |
| REQ_3.502 — 4-tier consumption reporting (Org / Team / User / Model) |
Intend to Comply |
4-tier budget enforcement ships (§6.2); matching report deck partial |
| REQ_3.503 — Unit Economics + Forecasting |
No Comply |
§43.1 |
| REQ_3.504 — Cost anomaly alerting (e.g., > 3× normal rate) |
No Comply |
§43.1 (threshold alerts ship; statistical anomaly detection does not) |
| REQ |
Disposition |
Evidence |
| REQ_3.601 — Visual map of Agents → MCP Servers → Data Sources |
No Comply |
§43.1 — Topology/Lineage Visualization |
| REQ |
Disposition |
Evidence |
| REQ_3.701 — User feedback (thumbs / stars) correlated with trace_id |
No Comply |
§43.1 — User Feedback System |
| REQ_3.702 — Visibility into when/why an agent capability was throttled |
Intend to Comply |
Events emit today (§22); dedicated dashboard view is a UI roadmap item |
| REQ |
Disposition |
Evidence |
| REQ_4.001 — HA failover, 99.99% uptime |
Intend to Comply |
§8, §31, §39 — architecturally supported; the 99.99% number is the customer’s deployment SLA |
| REQ_4.002 — Backup/restore w/ defined RPO/RTO |
Full Comply |
§31 |
| REQ_4.003 — Graceful degradation + self-healing |
Full Comply |
§8 |
| REQ_4.004 — Zero-downtime upgrades |
Full Comply |
§8, §31 |
| REQ |
Disposition |
Evidence |
| REQ_4.101 — Immediate global termination of AI agent activities |
Intend to Comply |
Manual org/user suspend + per-agent action API (§28); fleet-wide kill-switch automation is on roadmap |
| REQ_4.102 — Logging of termination events (cause, timestamp, affected agents) |
Full Comply |
§21 (audit trail) |
| REQ_4.103 — Auto-detect anomalies + termination from predefined policies |
No Comply |
§43.1 — Agent Kill-Switch Automation |
| REQ |
Disposition |
Evidence |
| REQ_4.201 — FCC + state laws + SOC 2 Type II, ISO 27001, GDPR, HIPAA |
Intend to Comply |
§20 — framework support ships; SOC 2 / ISO 27001 are customer-deployment attestations on the underlying infra |
| REQ_4.202 — Golden Security Configuration baseline |
Full Comply |
§20 (documented hardened preset) |
| REQ |
Disposition |
Evidence |
| REQ_4.301 — Integrate with observability, FinOps, Security tools |
Full Comply |
§21, §23 (OTLP + webhooks to ServiceNow / Jira / PagerDuty / Slack) |
| REQ |
Disposition |
Evidence |
| REQ_4.401 — AI-specific threat detection |
Intend to Comply |
DLP (§18), guardrails (§19), WAF (§24) detect AI-specific abuse; a unified “AI threat detection” pane is a UI roadmap item |
| REQ_4.402 — Integrate with existing SOC for investigation/triage/remediation |
Full Comply |
§23 |
| REQ |
Disposition |
Evidence |
| REQ_4.501 — Real-time PII / toxicity / injection filtering on data plane |
Full Comply |
§18, §19 |
| REQ_4.502 — Detect + block model outputs that mirror copyrighted / proprietary code |
Intend to Comply |
Webhook DLP plugin surface available (§18); first-party similarity engine not shipped |
| REQ |
Disposition |
Evidence |
| REQ_4.601 — Custom guardrails on request + response |
Full Comply |
§19 |
| REQ_4.602 — Integrate w/ 3rd-party PII, prompt-injection, content-moderation, groundedness tools |
Full Comply |
§19 (Bedrock Guardrails + plugin surface) |
| REQ_4.603 — Versioning + audit + rollback for guardrail policies |
No Comply |
§43.1 — Guardrail Versioning |
| REQ_4.604 — Guardrails on input prompts AND on tools/context data |
Full Comply |
§19 |
| REQ_4.605 — RBAC for guardrail policy management + audit-log observation |
Full Comply |
§17, §19 |
| REQ |
Disposition |
Evidence |
| REQ_4.701 — Canary + blue-green deployments w/ rollback |
No Comply |
§43.1 |
| REQ_4.702 — A/B testing for model/agent updates |
No Comply |
§43.1 |
| REQ_4.703 — Auto-forward test results to designated systems |
Intend to Comply |
Generic webhook forwarding available; dedicated reporter not native |
| REQ |
Disposition |
Evidence |
| REQ_4.801 — Lifecycle mgmt to deprecate/retire models + APIs to kill shadow callers |
Intend to Comply |
Models can be disabled (§4.2); deprecation pathing notifications and shadow-caller detection are roadmap |
| REQ_4.802 — Centralized registry of deployed models + versions + APIs w/ access enforcement |
Full Comply |
§4.2, §17 |
| REQ |
Disposition |
Evidence |
| REQ_4.901 — Alert when prompt/temperature changed in prod outside Control Plane |
No Comply |
§43.1 — Config Drift Monitoring |
| REQ_4.902 — Recompute baseline on demand |
No Comply |
(requires 4.901) |
| REQ_4.903 — Trigger alert within 1 monitoring window |
No Comply |
(requires 4.901) |
| REQ |
Disposition |
Evidence |
| REQ_4.1001 — Declarative policy language (e.g., Google CEL) |
No Comply |
§43.1 — Declarative Policy Engine |
| REQ_4.1002 — Per-agent tool whitelist enforcement |
No Comply |
§43.1 — Per-Agent Tool Whitelists |
| REQ_4.1003 — Granular access control (model / agent / MCP × user group) |
Full Comply |
§17 |
| REQ |
Disposition |
Evidence |
| REQ_4.1101 — On-prem self-hosted |
Full Comply |
§39 |
| REQ_4.1102 — Public cloud / managed cloud / multi-cloud / hybrid |
Full Comply |
§39 |
| REQ_4.1103 — Enterprise SLAs + strong isolation |
Intend to Comply |
§15 (multi-tenant isolation); SLAs are deployment-contract |
| REQ |
Disposition |
Evidence |
| REQ_4.1201 — Easy interface for deploying/updating gateway configs across environments |
Full Comply |
§34 (Gateway Mgmt UI) + import/export JSON |
| REQ_4.1202 — Configuration drift monitoring |
No Comply |
§43.1 |
| REQ_4.1203 — Reliable configuration rollback |
Intend to Comply |
Full-system rollback via .agwbak (§31); per-config-object rollback partial |
| REQ_4.1204 — Configuration validation tool aligned with vendor spec |
Full Comply |
§27 (AJV schema validation across routes) |
| REQ |
Disposition |
Evidence |
| REQ_4.1301 — GPU multitenancy + VRAM quota per NHI/app |
No Comply |
§43.2 (inference runtime responsibility) |
| REQ_4.1302 — Environmental + hardware telemetry (temp, power, PCIe) |
No Comply |
§43.2 |
| REQ_4.1303 — Local log aggregation (on-prem Splunk/ELK without cloud transit) |
Full Comply |
§22 (Kafka + air-gapped operation), §39 |
| REQ_4.1304 — Local image registry + provenance (Harbor, signed images) |
Full Comply |
§39 (any OCI registry; signatures are registry-policy enforced) |
| REQ_4.1305 — In-region model execution, no data exits the boundary |
Full Comply |
§39, §35 |
| REQ |
Disposition |
Evidence |
| REQ_4.1401 — HITL triggers w/ manual auth before agent action |
No Comply |
§43.1 |
| REQ_4.1402 — Log human reviewer identity + justification |
No Comply |
(requires 4.1401) |
| REQ |
Disposition |
Evidence |
| REQ_5.001 — Private VPC, no direct internet egress, explicit egress proxy w/ TLS inspection |
Full Comply |
§35, §5.3 (Zscaler integration), §29 |
| REQ_5.002 — WAF tuned for AI abuse (prompt flooding, token stuffing, abnormal session blocking) |
Full Comply |
§24 |
| REQ_5.003 — mTLS service mesh (Istio / App Mesh) |
Intend to Comply |
Agent mTLS ships (§28, §35); mesh orchestration is via deployment-layer Istio/Linkerd (§43.2) |
| REQ |
Disposition |
Evidence |
| REQ_5.101 — Okta as identity anchor for NHIs; registered scope + owning human + review cadence |
No Comply |
§43.1 — Okta NHI anchor model |
| REQ_5.102 — Permission intersection model (user × NHI scope) |
No Comply |
§43.1 |
| REQ_5.103 — Automated credential rotation + cert-bound tokens (RFC 8705) + PMO LLM oversight |
No Comply |
§43.1 — NHI Governance |
| REQ_5.104 — NHI lifecycle (creation / modification / decommissioning) + policies |
Intend to Comply |
Scoped API keys w/ TTL + rotation ship today (§35); formal NHI lifecycle model is roadmap |
| REQ |
Disposition |
Evidence |
| REQ_5.201 — Adversarial exercises + ongoing model validation |
No Comply |
§43.1 — Adversarial / Red-Team CI Framework |
| REQ_5.202 — Prompt injection resilience testing (direct + indirect / RAG-based) |
No Comply |
§43.1 |
| REQ_5.203 — Prompt drift analysis |
No Comply |
§43.1 |
| REQ_5.204 — OWASP LLM Top 10 + MITRE ATLAS testing |
No Comply |
§43.1 |
| REQ |
Disposition |
Evidence |
| REQ_5.301 — SSO via SAML / OIDC / Okta / Azure AD / Google Workspace + role sync |
Full Comply |
§16 (native SAML 2.0, OAuth2/OIDC, SCIM v2 role+group sync) |
| REQ_5.302 — Centralized API key mgmt w/ scoped + expiring keys |
Full Comply |
§16.1, §35 |
| REQ_5.303 — Fine-grained RBAC by user / team / server / tool / app |
Full Comply |
§17 |
| REQ_5.304 — Secure storage of secrets + credentials |
Full Comply |
§35 |
| REQ_5.305 — Audit log export to external systems + sensitive-info masking |
Full Comply |
§21, §22, §35 |
| REQ |
Disposition |
Evidence |
| REQ_5.401 — OAuth2 authorization code flow w/ PKCE |
Full Comply |
§16.3 |
| REQ_5.402 — Flexible OAuth2 (multi-IdP) |
Full Comply |
§16.3 |
| REQ_5.403 — Scope-based access control |
Full Comply |
§16.3 |
| REQ_5.404 — OAuth2 client credentials flow for M2M |
Intend to Comply |
M2M via scoped API keys ships today; strict OAuth2 client_credentials token endpoint is roadmap |
| REQ_5.405 — Swagger/OpenAPI specs for OAuth2 flows + scopes |
Full Comply |
§16.3, §25 |
| REQ |
Disposition |
Evidence |
| REQ_5.501 — API keys w/ expiration, scope to models/tools, auto-rotation + revocation |
Full Comply |
§16.1 (ak_… keys w/ TTL), §35 |
| REQ_5.502 — Unique NHI per agent w/ cryptographically bound creds propagated through call chain |
Intend to Comply |
Per-agent API keys + enrollment tokens (§28) ship; full NHI cert-binding (RFC 8705) is roadmap |
| REQ_5.503 — Permission intersection model (user role ∩ agent role, not union) |
No Comply |
§43.1 |
| REQ_5.504 — OpenAPI 3.x import + Developer Portal w/ try-it-out |
Intend to Comply |
OpenAPI generation + Swagger UI ship (§25, §16.3); arbitrary OpenAPI 3.x import for API-product cataloguing is roadmap |
| REQ |
Disposition |
Evidence |
| REQ_5.601 — Key rotation + revocation + integration w/ AWS Secrets Manager / Azure Key Vault / HashiCorp Vault |
Intend to Comply |
Rotation + revocation ship; external vault integration is roadmap (§43.1) |
| REQ |
Disposition |
Evidence |
| REQ_5.701 — Air-gapped / Dark Site deployment |
Full Comply |
§39 |
| REQ_5.702 — Model weight encryption at rest w/ HSM/TPM |
No Comply |
§43.2 (inference platform) |
| REQ_5.703 — Confidential Computing (NVIDIA CC / Intel TDX) |
No Comply |
§43.2 |
| REQ_5.704 — Hardware-backed identity (SPIFFE/SPIRE) |
No Comply |
§43.1 |
| REQ_5.705 — Micro-segmentation (SDN) |
No Comply |
§43.2 |
| REQ_5.706 — Local on-prem secret manager (Vault / CyberArk) integration |
No Comply |
§43.1 |
| REQ |
Disposition |
Evidence |
| REQ_6.001 — Playground for prompts / MCP / guardrails |
No Comply |
§43.1 — Prompt Playground |
| REQ_6.002 — Selectively enable MCP servers + tools in Playground (RBAC) |
No Comply |
§43.1 |
| REQ_6.003 — Code snippet generation (Python / Node / REST / LangChain / LangGraph) |
No Comply |
§43.1 |
| REQ_6.004 — Collaborate / share Playground experiments |
No Comply |
§43.1 |
| Line item |
Notes |
| Enterprise Platform License (Annual, base) |
Includes 1 control plane + 5 data planes; usage allowance to be specified at proposal time |
| Incremental Control Plane |
Per additional control plane |
| Incremental Data Plane |
Per additional data plane |
| Usage — Requests (per 1M) |
Tiered pricing above 250M requests; volume tiers in proposal |
| Premium Support — 24×7 |
Standard / Premium / Enterprise tiers |
| Implementation Services |
One-time integration fee |
| Training & Enablement |
Annual; covers admin + developer enablement |
| Control Plane on SaaS |
Optional vendor-hosted control plane |
| Category |
Full Comply |
Intend to Comply |
No Comply |
Total |
| Registry & Asset Management |
7 |
1 |
9 |
17 |
| Unified Gateway |
9 |
1 |
4 |
14 |
| Observability & Telemetry |
13 |
3 |
6 |
22 |
| Governance & Infrastructure |
14 |
5 |
13 |
32 |
| Security & Access Management |
13 |
4 |
16 |
33 |
| Sandbox & Developer Platform |
0 |
0 |
4 |
4 |
| Total |
56 |
14 |
52 |
122 |
Compliance rate: 46% Full Comply, 11% Intend to Comply, 43% No Comply.
Strongest categories (highest Full Comply density):
- Unified Gateway — 64% full comply; the LLM/MCP gateway is the platform’s core
- Observability & Telemetry — 59% full comply; OTel + Langfuse + Kafka + ClickHouse + Prometheus
- Security & Access Management — 39% full comply, with all 5 SSO + RBAC requirements green (REQ_5.301–5.305)
Weakest categories (highest No Comply density):
- Sandbox & Developer Platform — 100% No Comply; AgentWatch is not a playground product
- Registry & Asset Management — Agents + Prompts — Agent Registry, Prompt Registry, HITL workflow, and agent versioning are out of scope
- Hardware Security — TEE / HSM / SPIFFE / GPU multitenancy are infrastructure-layer concerns
Recommended positioning for enterprise RFP responses:
- Lead with the green columns: Unified Gateway, Observability, SSO/RBAC/SCIM, OpenAI-API compatibility, MCP proxy, DLP, Air-Gapped/Dark Site, and the on-device AgentWatch Connect agent (no direct competitor ships this).
- For Intend to Comply items, attach a target delivery date in the proposal’s roadmap column — see the dispositions above for the v2.0.63 baseline.
- For No Comply items, identify which are deal-breakers vs. nice-to-haves and propose integration partnerships (e.g., red-team CI via Lakera / Robust Intelligence, declarative policy via Open Policy Agent sidecar, NHI governance via Astrix / Entro, RAG quality via Ragas, prompt experimentation via Datadog LLM Experiments — see §45).
