A single agent is powerful. But the real magic begins when agents work together. Mothership is our answer to the question: "how do you get a team of AI agents to collaboratively execute complex tasks?"
Why one agent isn't enough
Imagine this task: write an analytical market report. You need to gather data from a dozen sources, analyze it, write the copy, fact-check it, and format it. In real life, one person doesn't do all of that. A team does: research analyst โ writer โ editor โ fact-checker.
Same goes for agents. One general-purpose agent that tries to do everything does everything mediocrely. Specialized agents working as a team deliver an order of magnitude higher quality.
The Mothership architecture
Mothership is the central orchestrator that manages a team of agents like a conductor leads an orchestra. It receives a task, decomposes it, distributes subtasks among agents, and assembles the results.
Task โ Mothership
โโโ Research Agent: data gatheringโ โโโ Perplexity + Firecrawl + Browser Use
โโโ Writer Agent: content creationโ โโโ GPT-4o / Claude Opus
โโโ Reviewer Agent: quality checkโ โโโ Evaluator + fact-checking tools
โโโ Final report assembly
Core principles
Parallelism by default. If two subtasks don't depend on each other, Mothership runs them simultaneously. One Research Agent gathers US market data while another analyzes the European market. The result: 3-5x reduction in execution time.
