what is impri fr?

Impri is a tool that puts a human checkpoint in front of actions an AI agent wants to take. Instead of letting an agent post a comment, send an email, or run a command right away, the agent proposes the action, it lands in an approval inbox, and a real person approves or rejects it before it actually runs. This gives teams a way to let AI agents work more independently while still keeping a person in control of anything risky or public facing. You can try it in two ways. There is a hosted cloud version where signing up through a simple web request or the app gives you an API key and an inbox URL in a few minutes, no installation needed. Or you can self host it with Docker Compose, which spins up a local server and web interface you open in your browser. A developer mode is also available for people who want to run the server and the web interface separately with live reloading while working on the code. Once set up, agents talk to Impri through a REST API or through an MCP server, which makes it easy to plug into tools like Claude Code. The API lets an agent push a new action, list pending actions, and check whether something was approved or rejected. There is also a command line tool for humans to review, approve, reject, and tail pending actions from a terminal, plus software development kits for Python and TypeScript, and integrations with tools such as LangChain, OpenAI Agents, CrewAI, n8n, Make, and Zapier. Impri also includes watchers, which are presets for monitoring places like Hacker News, Reddit, GitHub, npm, or arXiv, so an agent's proposed action can be triggered by outside events. Notification channels include Slack, Discord, Telegram, and email, so a human can be alerted when something needs a decision. The core project is self hostable for free under the MIT license, while the hosted cloud version and team features are the paid offering.