Durable execution, without replay.

Trigora is a durable execution substrate for long-running agents and dynamic software.

Its underlying execution architecture—Transparent Continuation Checkpointing (TCC)—preserves resumable program state at durable boundaries. Recovery from a committed continuation checkpoint does not require replaying the accumulated execution-history prefix.

import {
  effect,
  invoke,
  waitForEvent,
} from "@trigora/sdk";

export async function researchAgent(input: ResearchInput) {
  const sources = await effect(() =>
    searchWeb(input.query)
  );

  const analysis = await invoke(analyzeSources, {
    sources,
  });

  await waitForEvent("human.approved");

  return effect(() => publishReport(analysis));
}

Trigora is designed for programs that:

• call tools and external systems;
• wait for humans or events;
• invoke durable child executions;
• branch dynamically;
• survive for hours or days;
• recover after worker failure.

Cron, webhooks, queues, and API calls act as triggers that start durable executions rather than separate programming models.

Trigora is under active development.

The TCC research engine and evaluation prototype exist today. The public SDK, CLI, and managed Trigora Cloud platform are being built. Some code in this repository still reflects Trigora’s earlier event-execution product and should not be considered the final durable-execution API.

In a controlled evaluation at fixed live state:

• TCC recovery remained approximately 0.6–0.9 ms across history depths from 10 to 1,000.
• No semantic failures were observed across 50,000 generated cases within the tested TypeScript subset.

These are research-prototype measurements, not production performance guarantees.

Read the research · Technical report · Limitations

Website · Technology · Documentation · Design partners

Building a workload that needs durable execution? Contact omar@trigora.dev.