Taurus
my take on multi-agent hierarchies
Last checked:
What is Taurus?
Taurus is a multi-agent AI orchestration platform enabling users to deploy hierarchical teams of agents with persistent memory, isolated containers, and built-in collaboration tools. It supports scheduling, 30+ LLM models, and provides a web IDE with terminal access for agent coordination at scale.
Taurus pricing
Pricing model: Freemium
Free to try (no credit card required); paid plans available at taurusagents.com/plans
Taurus pros
- Hierarchical agent trees with persistent memory—each agent maintains its own knowledge base, filesystem, and context; agents can delegate tasks and coordinate through shared folders
- Real isolated Linux containers—full shell access, package installation, persistent storage, and browser capability; agents maintain state between runs
- Multi-model flexibility—supports 30+ models (Claude, GPT, Gemini, xAI, Groq, OpenRouter) usable interchangeably across teams
- Cron-based scheduling with memory continuity—agents can wake on schedule and pick up where they left off
Taurus cons
- Pricing details not displayed on main page—only 'free to try' visible; actual tier pricing requires visiting plans page
- No visible details on data residency, security certifications, compliance standards, or enterprise SLAs in main content
- No clear documentation of agent limits, monthly run quotas, storage limits, or resource constraints visible here
Frequently asked questions about Taurus
What is an agent on Taurus?
An agent is an AI entity with persistent memory, knowledge base, isolated Linux container, filesystem, and local repositories. Multiple agents form hierarchical trees for team-based orchestration.
How do agents collaborate in a tree?
Parent agents delegate tasks to children and review their work. All agents in a tree share a `/shared` folder for file exchange, repository collaboration, and inter-agent communication.
What models and LLM providers are supported?
Taurus supports 30+ models from Anthropic, OpenAI, Google, xAI, Groq, and OpenRouter. Different agents in the same team can use different models simultaneously.
Can agents run automatically?
Yes, via cron-based scheduling with overlap handling. Agents wake at specified times, access their memory, and resume previous work seamlessly.