Review before your agent acts
A browser-based review layer for autonomous AI agents. Agents propose actions, you inspect and edit in a real UI, then the agent continues.
$ npm install -g @harvenstar/agentclick
A browser-based review layer for autonomous AI agents. Agents propose actions, you inspect and edit in a real UI, then the agent continues.
Most agents go straight from decision to action. AgentClick adds a review step so you stay in control without slowing things down.
The agent sends an action, draft, or plan to AgentClick via a simple HTTP POST.
→A purpose-built UI opens in your browser. Read, edit, rewrite, or reject before anything runs.
→The agent picks up your edits and proceeds. Your preferences are saved for next time.
From emails to code diffs to multi-step plans, AgentClick gives you a real UI for every kind of agent interaction.
View your inbox, edit reply drafts paragraph-by-paragraph, and let the agent learn your writing style over time.
Visualize execution plans as a DAG. Add constraints, skip steps, or restructure before the agent starts.
See exactly which files change with full diff views before any command runs. File tree, add/modify/delete status, syntax highlighting.
Review tool calls, decisions, errors, and retries in a visual timeline. Catch mistakes and provide corrections.
High-risk actions get a dedicated approval screen with risk levels and space for notes before execution.
Track pending, active, and completed review sessions. Memory management, preferences, and session history in one place.
AgentClick is framework-agnostic. Any agent that can POST JSON and poll an HTTP endpoint can use it. Skill files make integration effortless.
Install globally, start the server, and tell your agent to use it.
# Install npm install -g @harvenstar/agentclick # Start the review server agentclick # For remote access (phone, another machine) agentclick --remote
Then tell your agent: "AgentClick is running at http://localhost:38173. Use it whenever you need a browser review UI."