Know what your agents did. Know what to fix.
Aletheia turns OpenAI agent runs into evidence-linked decisions, tool calls, customer-risk signals, and measurable outcomes—without exposing private chain-of-thought.
Customer cannot access Aster Cloud
usr_demo_42 · enterprise · EMEA
Verify enterprise customer
decision
Check service status · EMEA
mcp tool
Correlate outage with account
decision
Create priority support ticket
mcp tool
Escalate frustrated customer
outcome
Account-access cases in EMEA are escalating after a regional service degradation. Frustrated enterprise customers carry the highest risk.
Replay from “Service status checked” with the incident response playbook.
Follow every decision
Connect evidence-based decisions, model calls, MCP tools, state changes, retries, and outcomes in one trace.
Find customer risk early
Cluster support failures by issue, product, channel, region, sentiment, and escalation—not anecdotes.
Replay from the divergence
Fork the original state, change context or model settings, and preserve the complete lineage.
Measure agent economics
Compare latency, tokens, estimated model cost, retries, tool health, and resolution outcomes.
Instrument the OpenAI Agents SDK you already use.
Add one trace processor for complete model, tool, MCP, decision, duration, token, and cost telemetry. Sensitive content remains opt-in.
- Native OpenAI Agents lifecycle
- Evidence-based decision events
- Project-scoped, metadata-first capture
import { addTraceProcessor } from "@openai/agents";
import { createAletheiaTraceProcessor } from
"@avinash27/aletheia";
addTraceProcessor(createAletheiaTraceProcessor({
apiKey: process.env.ALETHEIA_API_KEY!,
openaiApiKey: process.env.OPENAI_API_KEY,
captureContent: false,
}));Observe behavior without collecting private reasoning.
Project-scoped keys, metadata-first capture, tenant isolation, bounded evidence analysis, and synthetic demo data by default.