The execution record for AI + distributed software

Ask your software why.

Causet records the causal chain behind decisions, events, state changes, and agent actions — so you can ask why, follow the evidence, and replay what happened.

Explore an execution Run locally
HTTP·Webhooks·Kafka·CDC·EventBridge·SDK
Ask Causet answering why a payment was blocked with verified causal evidence
Real Causet Cloud demo · verified causal evidence
The product

Every outcome comes with evidence.

Ask why, follow the recorded chain, then rewind to the state that existed at any point in the run.

Ask Causet

Answers grounded in the run.

Answers cite the decisions, inputs, events, and state from the actual run.

✓ Cited execution evidence
Ask Causet answering why a payment was blocked with a verified summary, state details, and causal chain
“Why was this payment declined?” “What changed this transaction’s state?” “What happened before this decision?”
Why Causet

Your evidence is scattered.
Causet records the chain.

Logs and traces show pieces. Causet preserves the application-level relationship between input, decision, state change, and effect.

Today
LogsTracesKafkaDatabaseAgent tracesWebhooks
?Engineer reconstructs
what happened
With Causet
Observationexternal event received
↓
Decisionpolicy / AI / rule executed
↓
State changefields committed
↓
Effectevent / projection / callback
One causal execution record
01
↳
Observe

Ingest an event, webhook, API call, CDC change, or agent action.

02
◆
Decide

Record the policy, rule, deterministic transition, or AI decision that ran.

03
Δ
Change state

Commit the state transition with the business events it produced.

04
✓
Connect effects

Link projections, callbacks, tool calls, and downstream events to their cause.

05
✦
Explain

Inspect, query, fork, rebuild, or replay from the recorded evidence.

Ask why·Trace state·Replay / fork·Audit evidence
What this looks like

When “what happened?” is not enough.

Three situations where reconstructing the story from telemetry is not good enough.

AI agent

Your agent refunded $2,800.
Why?

See the state, decision, tool call, and downstream change as one execution.

context→decision→tool→state
Payments

A payment settled with the wrong status.
Where did it diverge?

Trace authorization, callbacks, ledger events, and state as one chain.

authorized→callback→ledger→state
Production incident

Customer cust_81 is suspended.
What caused it?

Start from current state and walk backward to the decision and input that caused it.

state←event←decision←input
Fits your stack

Start with one execution path.

Keep your APIs, database, queues, and observability. Send Causet the flow you need to explain.

HTTPWebhooksKafkaCDCEventBridgeSDK
existing-route.ts
// Keep your route. Hand one critical transition to Causet.
const result = await client.submitIntent(
  'refund_stream',
  refund.id,
  'REFUND_REQUESTED',
  { refund_id, order_id, amount, reason },
  refund.id // idempotency key
);

return Response.json(
  { status: 'accepted' },
  { status: 202 }
);
execution record created run_8df4…
Causet forks view showing a healthy production fork and controls to clone state into another environment
Before you touch production

Fork the state. Test another path.

Clone ledger state into an isolated fork, deploy or replay there, and compare the result without mutating main.

See forks and replay →
Built, not mocked

Run it. Inspect it. Decide for yourself.

Install the CLI, run locally, inspect public releases, or read the SDK source.

causet local upLocal runtime · no cloud login
Official installer→CLI releases v10.0.12→SDK source MIT→Documentation→
Start with one flow

Give your software an answer to “why?”

Run Causet locally