tool-permissions

Open-core runtime for Agentic Control Plane: identity, policy, limits, routing, PII redaction, and audit as npm modules.

One-command install for Agentic Control Plane — Claude Code, Cursor, Codex, OpenClaw, MCP. Policy-checked from call one.

48-scenario benchmark testing identity, policy enforcement, and observability across AI agent runtimes.

Composable autonomy primitives for Vercel AI SDK — permissions, cost tracking, interrupts.

Agentic Control Plane for Meta Muse Code — policy-check every tool call before it runs

Hermes Agent plugin for Agentic Control Plane: policy-check, audit, and veto every tool call.

OpenClaw plugin for Agentic Control Plane: policy-check, price, and log every tool call.

Codex CLI plugin for Agentic Control Plane: policy-check, price, and log every shell command.

TypeScript + Python SDKs for scope-narrowed subagents and delegation chains — every action audits back to the originating human. Anthropic SDK, CrewAI, LangChain.

Grok Build (xAI) hook for Agentic Control Plane: policy on every tool call, even in always-approve mode.

Claude Code plugin for Agentic Control Plane: policy-check, price, and log every tool call. One-command install.

Regex tool-permission gate for omp: per-tool rules over tool arguments, a project_root boundary, and symlink-escape detection. Ports the decision semantics of Zed's agent.tool_permissions.

Reference implementation for securing agentic AI apps with guardrails, tool permissions, and audit logs.

Agentic Control Plane hook for Google Antigravity (agy) — policy on every tool call, ACP asks as native force_ask prompt cards

Public-safe agentic engineering workflow patterns covering execution loops, tool permissions, review gates, logging, refusal rules, and human approval.

Prime Agent extension for Agentic Control Plane: policy-check every tool call — no built-in permission framework.

MCP server for Agentic Control Plane: policy-check every tool call from Claude, ChatGPT, Cursor, and other MCP clients.