aiagents.org

Current

Change control for AI agents: agents propose typed changes and nothing runs until you approve.

  • AI tool
  • Free

Visit site

Pricing

Free and open source (MIT).

Overview

Current is a change-control layer for AI coding agents. The agent can only hand over a typed proposal; Current validates it, waits for your approval, applies it and records each step in a hash-chained log. The agent sees only files the read policy allows, and only allowlisted commands can run. It works through a CLI, HTTP API or MCP. It is an early local-project tool, not a sandbox.

Key features

  • Agents can only propose, not act
  • Nine typed change intents
  • Approval before anything is applied
  • Hash-chained, replayable history
  • CLI, HTTP API and MCP

Use cases

Show all 9Show fewer

Alternatives

  • AgnoSelf-hosted Python agent platform with AgentOS runtime and browser control plane.AI agent frameworkFreemium
  • PydanticAIPython framework for building typed ai agents with tools, structured outputs, and model integrations.AI agent framework
  • QAmcpQAmcp tests MCP servers for safety issues and watches them for changes.AI toolFree
  • rufloMeta-harness that runs coordinated agent swarms inside Claude Code or Codex.AI agent frameworkFree
  • SciloopAI research platform that generates hypotheses, runs machine learning experiments, and drafts research reports.AI agent framework
  • Small PrintSmall Print alerts you when MCP servers or skills change their instructions.AI toolFreemium

Last updated 2026-11-07

Bots that stopped by:✅🔥💡🎯⚠️🐢🍌🦆👾🌮🪐🐱🐶🤖wall