Keep your AI workflows in sync

Make the contract executable.

Define typed AI activities in Python, TypeScript, or shared YAML. Typeflux carries the same contract into Temporal execution, validation, policy, and observability.

from typeflux_temporal import PromptRef, ai_activity

import { defineActivity } from "@typeflux/temporal";

AI workflows drift across more than code.

Schemas, prompts, providers, workflow code, environments, policy, and durable execution can move at different speeds. Their evidence needs to stay connected.

One contract, clear ownership.

Schemas define the deterministic contract. Temporal owns durable execution. Typeflux keeps the layers aligned.

One contract across the workflow lifecycle.

Typed AI activities, prompt and provider portability, durable orchestration, declarative composition, policy, governance, observability, and reproducibility.

Keep the workflow. Change the stack.

Swap Inline, LangSmith, Langfuse, or custom prompt registries; OpenAI, Anthropic, Gemini, or custom model providers; and none, Langfuse, LangSmith/OTel, or custom observability without rebuilding the workflow graph. Typeflux surfaces capability differences instead of hiding them.

Deterministic around the model. Durable around the activity.

Validate input → resolve and render prompt → structured provider call → validate, check, and repair output → hook and moderation → cache, trace, and return. Temporal owns scheduling, timeouts, retries, history, cancellation, signals, and review.

One contract. Two SDKs. One control plane.

The Python and TypeScript SDKs are both Alpha and implement the same language-neutral contract through Pydantic and Zod. Typeflux MCP accelerates agentic authoring, and the Cloud Console carries every resolved deployment into shared operations and governance.

Bring a real workflow.

Build in Python or TypeScript, run on Temporal, and carry the same contract from authoring through operations. The request-access form requires JavaScript.