awaithumans

Overview

Open source human-in-the-loop library for production LLM agents. It pauses an AI agent to ask a human via Slack, email, or web dashboard, then resumes with the answer. Includes typed Pydantic and Zod responses, durable Temporal and LangGraph adapters, an AI verifier, and an audit trail. Self-hosted under Apache 2.0, available in Python and TypeScript.

Best for

Best for
Teams building production LLM agents on Temporal or LangGraph that need a self-hosted human approval step.

Use cases

• Pause an agent for human approval before executing a critical action
• Route questions to a human via Slack or email from within an agent workflow
• Maintain an audit trail of human interventions for compliance

Notes

Open source human-in-the-loop library for production LLM agents. It pauses an AI agent to ask a human via Slack, email, or web dashboard, then resumes with the answer. Includes typed Pydantic and Zod responses, durable Temporal and LangGraph adapters, an AI verifier, and an audit trail. Self-hosted under Apache 2.0, available in Python and TypeScript.

24 stars on GitHub. Last updated 2026-09-11. Licensed Apache-2.0.

Use cases

• Pause an agent for human approval before executing a critical action
• Route questions to a human via Slack or email from within an agent workflow
• Maintain an audit trail of human interventions for compliance

Pros

• Open source and self-hosted, giving full data control
• Typed responses in Python and TypeScript reduce integration errors
• Includes durable adapters for Temporal and LangGraph

Cons

• Small community (24 stars) means limited adoption and support
• Requires self-hosting and ongoing operational maintenance
• Tightly coupled to Temporal and LangGraph, less useful for other agent frameworks

Indexed from awesome-llmops and enriched against its public facts.