कामCREW: a working home for an AI team.

01

The problem

Agent work can be difficult to follow when task status, conversations, configuration, and research live in separate places. A person needs enough context to understand the state of the work and decide when to intervene.

02

My contribution

I built the Next.js and TypeScript control interface and the server routes that connect it to the Hermes gateway and task state. The workspace brings the task board, agent conversations, model configuration, and research workflows together.

Hermes provides the agent runtime. I did not invent that runtime; my contribution is the control center built around it.

03

How the pieces connect

• Next.js and TypeScript provide the interface and control layer.
• Server routes connect to the Hermes gateway.
• SQLite-backed state supports the task board.
• Server-sent events bring live work updates into the interface.
• Pause and resume controls give the operator a way to intervene.

04

What makes the interface useful

The goal is to show what is actually happening. Tasks and conversations can be reviewed together, connection failures are represented as failures, and model settings are accessible in the same workspace.

This makes the control boundaries visible: the operator uses the interface, the server routes connect to state and the gateway, and Hermes performs the underlying agent work.

05

Current scope and limitations

This is an independent engineering project, not an established commercial SaaS case study. No paying-customer count, revenue, measured time savings, or public live-demo URL is claimed.

DeepSeek, vector-database retrieval, MCP, and agent evaluations are areas I am exploring. They should not be read as completed कामCREW capabilities unless a later implementation documents them.