Brin: The Agent That Connects Everything

I have too many tools.

Mesh for memory. Rein for workflows. Surfer for browser automation. Deck for virtual desktops. A dozen MCP servers. 52 specialist agents. 23 workflow templates.

Each one works great on its own. Together, they're a mess. Which tool handles this request? Which specialist should I ask? Where do I even start?

Brin is the answer to "I don't want to remember which tool does what."

What Brin does

Brin is a unified router. You talk to Brin, and Brin figures out which tool, specialist, or workflow to call.

Ask a question about code architecture? Brin routes it to the architect specialist. Want to search your memory? Brin calls Mesh. Need to run a multi-agent deliberation? Brin starts a Rein workflow. Want to take a screenshot of a website? Brin talks to Surfer.

One interface for everything.

How it works

Brin is an MCP server that aggregates all my other services:

• 52 specialists -- AI agents with defined expertise (critic, researcher, architect, etc.)
• 23 workflows -- Rein workflow templates (deliberation, code review, brainstorm)
• Mesh memory -- semantic search, save, tag operations
• Surfer -- browser automation (fetch, screenshot, headless interaction)
• Diagnostics -- server health, process monitoring

Each capability is exposed as an MCP tool. Claude Code, Cursor, or any MCP client can call them.

The specialist layer

Every specialist is a Markdown file with expertise and constraints:

architect-specialist.md
critic-specialist.md
researcher-specialist.md
security-specialist.md
...52 files

When you ask Brin a question, you can target a specific specialist:

"Ask the security specialist: is this auth flow safe?"

The specialist gets the question plus its expertise context. The response comes from a model that's been primed with deep domain knowledge for that specific role.

Why not just use Claude directly?

You can. And for simple questions, you should.

Brin is useful when:

• You need domain expertise that benefits from a detailed system prompt (security, architecture, UX)
• You want to search across services -- memory + docs + browser in one query
• You want to start a workflow -- something that takes multiple agents and 5 minutes to run
• You're working in an MCP-compatible editor and want everything accessible as tools

It's not about replacing direct LLM access. It's about reducing the cognitive overhead of remembering which tool does what.

Agent Spaces

Different agents need different contexts. The content manager shouldn't see infrastructure worklogs. The security auditor doesn't need marketing notes.

Brin supports workspaces -- isolated views of memory, different specialist sets, different default workflows. Each agent gets its own space.

The compound effect

The interesting thing about Brin is not any single feature. It's the compound effect of having everything connected.

A question like "what did the deliberation team decide about the auth approach last week?" requires:

1. Understanding the intent (search memory)
2. Knowing the right tags (agent:team-deliberation, topic:auth)
3. Searching semantically (meaning-based, not keyword)

Brin does all three in one call because it knows about Mesh, knows the tag schema, and knows how to route.

No single tool does this. The glue is the product.