Tale

Tale is a shared project workspace where teams can assign work to people or AI agents, provide the relevant files and instructions, and review the resulting reports, code, and other deliverables. It brings task boards, conversations, shared knowledge, agent configuration, automations, approvals, and audit records into one interface.

Agents run inside persistent sandbox workspaces with configured resource limits, tools, models, and repository access. Supported runtimes include Claude Code, Codex, Gemini, OpenCode, Hermes, OpenClaw, Cursor, and Pi. Teams can connect external services, compare model responses, build recurring workflows, and require approval before selected actions proceed.

The Community edition includes the full product under the MIT license. Self hosting uses a seven container stack with Docker, while Kubernetes deployment is documented for more advanced installations. You control the application data and infrastructure, but cloud models, connectors, and external tools can still send information outside your network. Model usage and hosting costs are separate.

Tale has detailed guidance covering secrets, encrypted provider credentials, sandbox tokens, network restrictions, TLS, backups, and production hardening. Even so, agent execution carries meaningful risk. Sandboxes may execute code and access repositories or connected services, while optional nested Docker changes the isolation requirements. Tale is best suited to technical teams prepared to operate and secure a complex application rather than users looking for a simple task manager.