Agents

Three layers — workspace, project, and daemon. All inherit your brand DNA.

Every Clearly workspace has agents. Agents are not chatbots — they are scoped context-aware processes that read your folders, call your tools, and write back. They inherit your brand DNA so they sound like you out of the box.

Three scopes

Agents come in three scopes, from broadest to narrowest:

1

Workspace agent (Omni)

One agent per workspace, knows everything: every project, every doc, every message. Best for "what was decided about X?" or "summarize this week."

2

Project agents

Pro+: one agent per project. Scoped to that project's files, channels, and members. Best for project-specific work that shouldn't leak across teams.

3

Daemon agents

Power users only: agents that run on your local Mac via the Beehaven daemon. Can call your terminal, your apps, your filesystem. The bridge between your workspace and your computer.

MCP tools

Every workspace exposes an MCP server. Connect Claude Desktop, Cursor, or any MCP-compatible client and your agents become callable as tools. Project context, brand DNA, and folders all flow through. Free on every plan.

Claude Desktop config

json
{
  "mcpServers": {
    "clearly": {
      "command": "beehaven",
      "args": ["mcp"]
    }
  }
}
Pro tip: Tell your project agent your style preferences once, in writing, and it sticks. The brand DNA wizard already captures voice + tone, but per-project quirks ("never use "premium," prefer "considered\"") lock in fast.