Why one AI can't do every job
Coding, writing and research call for different instructions, tools and context. In NanoCrew, we organize that work as separate agents with their own identities, memories and workspaces.
The design choice is how to organize the work around the model.
Separate the role, context and tools
A crew lets you make those responsibilities explicit:
- Role. Give each agent a specific identity and job.
- Context. Keep its workspace and long-term memory with that role.
- Tools. Choose the tools the agent needs, and review sensitive actions.
This is an architectural choice, not a claim that a general-purpose assistant cannot use memory, tools or scheduled tasks. Read the self-hosting considerations.
A crew of specialists
Here are some roles you might configure, depending on the tools and models you provide:
- A dev agent that ships features, with your repo in its memory.
- A content agent that writes and creates in your voice.
- A research agent that reads, summarizes, and tracks topics over time.
Each one has its own skills, its own tools, and its own memory. They can hand work to each other and run from chat, schedules or webhooks while your deployment is running.
The current release is early. Review its setup and limits before relying on it for a workflow.
That team is exactly what we built with NanoCrew in Open source. One platform, a specialist for every job — reachable from your dashboard and WhatsApp.