Bloome

Bloome is a chat-based workspace where people and AI agents can participate in the same conversation. Its features page describes agents that share context, delegate and review work together, with threads and mentions connecting them to the team. Multiple agents can help organize a task, but agreement between generated answers should not be treated as proof that a claim is correct.

Give the chat a defined task

The publisher presents workflows for researching, drafting and revising content. It also lists integrations for coding agents such as Claude Code, Codex and Gemini CLI. Check the setup requirements and available permissions for the particular connection before bringing it into a shared conversation.

Make the source material and acceptance criteria clear. If an agent drafts a customer-facing document, a human still needs to verify factual claims and decide whether it is ready to send. Keep confidential material limited to the people and agents that need access rather than assuming every participant has the same role.

Clone an agent or install a skill

Bloome's agent marketplace is an Explore library of public agents. The publisher says users can clone an agent, customize its prompt and tools, add it to a chat or publish their own. Inspect the clone's configuration instead of treating its public availability as an assurance of quality.

The separate skill marketplace uses SKILL.md bundles that teach a specialized capability. Skills can be installed from a chat card, a public URL or a ZIP upload. An agent and a skill are different objects: cloning a participant is not the same action as adding instructions to an agent you already own. Review an external bundle before granting tools or access to private information.

Check cloud access and operating limits

The features page labels cloud agents as beta and says they can stay reachable when your laptop is closed. Confirm what runs in that mode and how you can stop or inspect it before relying on an unattended task.

The feature descriptions do not provide a complete numeric price schedule or establish identical access across every integration. Check the account's actual allowances and the applicable handling of shared conversation data. Start with a bounded, nonsensitive task and inspect the work and permissions before expanding the agent's responsibilities.