Causal system intelligence

The system graph
for distributed software.

Provenance models your system's behavior as a causal graph. Every interaction linked to what caused it and what it triggered — across services, queues, and time.

Free tier includedNo credit cardNode · Python · Java · TSOpenTelemetry compatible

The problem

Your systems talk.
Nobody's listening.

Microservices, queues, webhooks, Lambdas — your operations span dozens of systems. When something breaks, you jump between five tools trying to piece together what happened.

📋

Logs show what — never why

🔗

No link between related events across services

Alerts fire after users already noticed

🔍

Debugging means 5 tools, 1 hour, 3 engineers

🤖

AI can't reason about unstructured noise

📊

Business teams can't self-serve answers

The insight

Systems aren't logs.
They're causal graphs.

Every action in your system causes other actions. A payment triggers inventory, which triggers shipping, which triggers a notification. That's not a log line — it's a graph of cause and effect.

Interaction

The atomic unit. Every time something happens — a user acts, a service calls another, a webhook fires — that's an interaction. Structured, typed, and linked to what caused it.

Unit of Work

The full business process. Order-to-cash. Signup-to-activation. Payment-to-delivery. One ID links every interaction in the chain — across services, queues, and time.

One interaction → full causal chain

Order CreatedPayment CapturedInventory ReservedShippedCustomer Notified

5 services. 1 Unit of Work. Full causal reconstruction.

A new generation

Beyond observability.

Logs gave us data. Dashboards gave us visibility. Provenance gives us understanding.

1Gen 1 — Logs

Raw text output. Manual grep. No structure. No correlation. Debugging by reading thousands of lines.

2Gen 2 — Observability

Unified dashboards. Metrics, traces, logs in one place. Better — but still fragmented at the model level. You see symptoms, not causes.

3Gen 3 — Provenance

Native causal graph. Business-process-aware. Every interaction linked to its cause and effect. Full system behavior reconstruction from a single ID.

Causal modelUOW trackingAuto-reactionsAI-ready

How it works

One order. Nine services. One graph.

An order flows through payment, fraud, inventory, shipping, and notifications. Provenance captures the full causal chain as a single Unit of Work.

Order-to-Cash LifecycleUOW: 7f3a9c2d
Order Createdweb-app
Payment Capturedstripe-webhook
Fraud Check Passedfraud-svc
Inventory Reservedwarehouse-api
Shipment Requestedlogistics-api
Customer Notified→ SendGrid
Ops Alert→ Slack #orders
Deliveredcarrier-webhook
Invoice Generated→ AWS Lambda
System event Auto-reaction

Capabilities

Everything connects.

🌳

Causal Event Graphs

Every interaction links to its cause and its effects. Navigate the full chain — not isolated events.

🔗

Unit of Work Tracking

One ID connects an entire business process across services, queues, and time. From trigger to consequence.

Automatic Reactions

When something happens, fire Slack messages, emails, webhooks, or Lambda functions. No polling. No cron.

🐛

Error Tracking

Capture exceptions with stack traces, grouped by fingerprint. Each error links to the full workflow that caused it.

🧠

AI-Ready System Memory

Structured causal history that AI agents can query, reason about, and act on. Built for RAG and MCP.

📡

OpenTelemetry Ingestion

Use alongside your existing tools. Ingest spans from any OTel-compatible system into the causal graph.

Reactive by default

Observe. Then act.

Provenance doesn't just record what happened. It reacts. Define subscriptions — when a condition is met, the system responds automatically.

💬

Slack

Post to channels on any event

📧

Email

SendGrid templates with dynamic data

🔗

Webhooks

Hit any URL with the full payload

λ

Lambda

Run custom logic on every trigger

Reactions are part of the causal graph — they appear as child interactions in the Unit of Work tree.

AI-native

System memory for AI agents.

Logs are noise to LLMs. Provenance gives AI agents structured, causal, queryable system history — ready for reasoning.

MCP Server

AI agents query your system history, investigate incidents, and create configurations using natural language.

Structured for RAG

Every interaction is typed, timestamped, and causally linked — ideal for retrieval-augmented generation.

Causal Reasoning

Trace any outcome back through the chain of events that caused it. Full causality for AI decision-making.

Operational Memory

Your system's complete behavioral history in a format AI can understand, search, and act on.

Integrate in minutes

One SDK call. You're live.

SDKs

Node.js

npm i @stdiolabs/provenance-sdk

Python

pip install provenance-sdk

Java

provenance-sdk-java

TypeScript

npm i @stdiolabs/provenance-sdk-ts

CLI

$ provenance track \
    -r order-123 \
    -t ORDER \
    -a CREATED

✔ Interaction recorded (12ms)

$ provenance trace --uow 7f3a9c2d
  ORDER/CREATED    web-app
  PAYMENT/CAPTURED stripe
  INVENTORY/RSRVD  warehouse
  SHIPMENT/SENT    logistics

Works With

  • OpenTelemetry exporter
  • Auto-instrumentation (Express, FastAPI, Spring)
  • Inbound webhooks (Stripe, GitHub, etc.)
  • Slack, Email, Discord reactions
  • AWS Lambda triggers
  • MCP server for AI agents

Beta Access

Free during Closed Beta

Full access to every feature. No credit card. No catch.

One SDK call. A complete system story.

Start building your causal graph today. Free tier included.

No spam. We'll only email you when your spot is ready.

No credit card required. Free tier forever.