Integrations
Approval gates for every agent framework
Your framework can pause a run. Pliuz is the production approval layer behind the pause — where the request is stored, how it reaches a human, and how every decision is audited — added with a single wrapper.
Pick your framework
LangGraph's interrupt() pauses a run, but leaves storage, routing, and audit to you. Pliuz adds production approval gates with a tamper-evident audit trail.
Add human-in-the-loop approval to OpenAI Agents SDK tool calls. needs_approval pauses the run; Pliuz adds Slack/web approvals and a tamper-evident audit trail.
CrewAI's human_input=True pauses a run on the console. Pliuz adds the production approval layer: Slack/web inbox, one-line wrapper, tamper-evident audit trail.
Claude Agent SDK pauses on canUseTool, but leaves storage, routing, and audit to you. Pliuz is the production approval layer: Slack/web inbox + tamper-evident trail.
Vercel AI SDK gives you needsApproval to pause a tool call, but leaves storage, routing, and audit to you. Pliuz is the production approval layer that fills the gap.
Why a separate layer at all
Every modern agent framework can pause a run for a human — LangGraph with interrupt(), the OpenAI Agents SDK with needs_approval, the Claude Agent SDK with canUseTool, the Vercel AI SDK with needsApproval. None of them answer where the request is stored, how it reaches the right person, or how the decision is proven later. That is the production gap Pliuz fills, and it is the same gap whether you run one framework or several.