Agent registry
Prometa AI Operating System tracks ownership, status, environments, connected tools, related capabilities, workflow usage, and Prometa reference templates in one operational layer.
Coordinates recommendation, pricing, and campaign tools across retail growth workflows.
Owner usr-004 · latest version v2.3
Last execution 2026-03-16T12:05:00Z · 14.8k orchestrations / month
Production-tested by Prometa
Classifies intake packets, invokes extraction tools, and routes claims into reviewer queues.
Owner usr-003 · latest version v1.6
Last execution 2026-03-16T10:18:00Z · 2.4k packets / month
Pilot-ready
Runs scenario-led forecasting checks and escalates anomalies to planners.
Owner usr-001 · latest version v1.2
Last execution 2026-03-16T09:44:00Z · 640 scenario runs / month
Observes integration health, policy breaches, and release readiness across tenant environments.
Owner usr-004 · latest version v3.1
Last execution 2026-03-16T12:11:00Z · 96 control checks / hour
Actively used by Prometa
Runtime engine
Normalizes input payloads and request context before execution.
payload validation + request shape
Selects allowed tools and routes calls by policy, env, and confidence state.
tool route + dependency health
Injects session, long-term, and vector context according to retention policy.
context source + retrieval latency
Builds final output, action bundle, and escalation path.
output quality + latency
Runs pre and post execution safety, data, and policy checks.
policy hits + blocked actions
Configuration graph
Optimize revenue outcomes while respecting policy and promotion constraints.
Overrides: per request policy override · per environment channel restriction
Route intake packets into review-safe claims paths with document context.
Overrides: pilot-only reviewer requirement · per request masking policy
Multi-environment runtime
Free execution environment
Promotion path dev → test → pilot → prod · 0 execution contexts
Protected environment
Promotion path dev → test → pilot → prod · 1 execution contexts
Protected environment
Promotion path dev → test → pilot → prod · 1 execution contexts
Protected environment
Promotion path dev → test → pilot → prod · 1 execution contexts
Version operations
Added campaign throttling threshold and reviewer checkpoint for margin-sensitive segments.
prompts, routing logic, policy gates, fallback tool order
Created by Prometa Admin on 2026-03-15
Expanded extraction policy and confidence escalation path for missing document bundles.
tool permissions, confidence thresholds, human review routing
Created by Aylin Demir on 2026-03-12
Enabled regression suite against golden demand scenarios before planner release.
forecast inputs, benchmark set, reviewer scoring
Created by Elif Kara on 2026-03-10
Execution trace
Trace `/trace-001` exposes orchestration steps, tool calls, decision points, latency, token cost, and suggested remediation without opening backend internals.
Scenario pack normalized
Forecast API
Peak cluster anomaly detected
1240 in · 318 out · $2.10
model 420 ms · tool 980 ms · orchestration 210 ms
Escalated to planner queue
Planner Review UI
Confidence below threshold
220 in · 91 out · $0.60
model 180 ms · tool 220 ms · orchestration 140 ms
Decision path stayed inside registered tools, applied runtime guardrails, and routed to review whenever confidence or policy state required intervention.
Planner scenario payload missing one required pricing field. Suggested fix: Add schema validation in input handler and inject default pricing floor before review.