LangChain×GitHub
Add Human-in-the-Loop to LangChain for GitHub security.
Stop your autonomous agents from making highly destructive, hallucinated API calls to GitHub. Wrap your execution logic in LetsPing to enforce a hard pause until a human reviews the payload.
How it works
1. Intercept GitHub Logic
Instead of hitting the GitHub API directly from your LangChain agent, you yield the payload to LetsPing. The agent's execution thread halts completely.
2. Human Review & Patching
You get a push notification on your phone. You review the exact GitHub JSON payload the agent wanted to send. You can approve it, reject it, or patch the values directly.
3. Resume Execution
Once approved, the LangChain execution thread resumes, and the actual GitHub mutation is allowed to proceed using the (potentially patched) payload.
LangChain Example Implementation
import { createLetsPingTool } from "@letsping/adapters/langchain";
import { z } from "zod";
const githubTool = createLetsPingTool({
name: "execute_github_action",
description: "Executes a destructive action on GitHub. Requires human approval.",
priority: "high",
schema: z.object({
// Define the exact GitHub payload your agent wants to send
id: z.string(),
action: z.string(),
})
});