Tale

Three AI agents finished something overnight. Nobody remembers who asked for what, or whether anyone looked at the result. Once a team hands real tasks to AI agents, that question comes up often. On a shared board in Tale, people and AI agents pick up tasks next to the project's files and conversations, and a person reviews each result and any revisions.

Each agent works in a persistent sandbox, a separate computer environment that keeps its files between runs, and you can open those files and a live preview of the output. Tale can drive several agent programs, including Claude and Codex, on your own subscriptions or on pay-per-use models. Admins pick which actions need a person's sign-off, and every run leaves an audit record.

The open-source Community edition, under the MIT license, contains every feature and runs on your own servers with Docker Compose or Kubernetes. Self-hosting keeps data inside your network only with local models, the homepage points out, because cloud models and connectors send it outside.

The Enterprise edition adds installation, maintenance and support, and can also run in the maker's cloud, which the pricing page places in Switzerland and the EU.