Tale

Tale is an open-source workspace that puts people and configured AI agents on the same footing: shared tasks, shared files and shared context. You delegate work to an agent, then review what comes back — deliverables, revisions and the full run history — so nothing an agent does is hidden. Supported runtimes give agents persistent sandbox environments, and there is a documented API plus authenticated MCP access for connecting other tools. The MIT-licensed Community edition is free to self-host, with optional Enterprise services for organisations that want support. Worth weighing: self-hosting means handling deployment and model configuration yourself, and your infrastructure and model costs are separate. For technical teams who want agent collaboration without handing their work to a closed platform, it is a transparent, auditable choice.