What AI models does the agent use?

The agent has access to Claude, DeepSeek, Gemini, and other top models. It automatically routes to the best model for each task — you never have to think about API keys or token limits.

Yes. Each agent runs in an isolated environment. We don't train on your data, and you can export or delete everything at any time. The underlying engine (GoGogot) is open-source, so you can audit exactly how it works.

What can the agent actually do?

Browse the web, process files (CSV, PDF, images), run scheduled tasks, remember context across sessions, and communicate via Telegram or Slack. Think of it as a capable junior employee, not a chatbot.

How does memory work?

The agent maintains persistent memory across sessions. It remembers your preferences, past conversations, and task context. You can also explicitly tell it to remember or forget things.

Can I self-host the agent?

Absolutely. GoGogot is 100% open-source. Run it on your own hardware for free. cowork.ink is the managed version — we handle servers, model costs, updates, and uptime so you don't have to.

Is there a free trial?

We offer a 7-day money-back guarantee. Spin up an agent, give it real tasks, and if it doesn't save you time, we'll refund you — no questions asked.

What roles should an AI agent team have?

Every effective AI agent team needs at least an orchestrator (the planner that decomposes tasks), one or more specialist workers (domain-focused executors), and a verifier (critic agent that checks outputs). Larger teams also benefit from a router agent for intent classification and a memory/state agent for shared context. See our [guide to types of AI agents](/blog/types-of-ai-agents/) for definitions of each.

What is the difference between an orchestrator agent and a worker agent?

An orchestrator agent handles high-level planning — it breaks a complex task into subtasks and delegates them to worker agents. Worker agents are specialists that execute one well-defined capability (e.g., web search, code generation, data analysis). Orchestrators synthesize results; workers produce them. The division is analogous to a project manager and individual contributors on a software team.

How many AI agents should be on a team?

Start with 3–5 agents for most tasks. Anthropic's production multi-agent research system runs 3–5 subagents in parallel. Beyond 7–8 agents, coordination overhead typically outweighs the parallelization benefit. The right number depends on task parallelizability, not on "more is better." Add agents only when hitting a concrete operational limit — context window exhaustion, speed, or specialist knowledge gaps.

What communication topology is best for AI agent teams?

It depends on your task structure. Hierarchical (orchestrator-worker) is best for complex tasks requiring central coordination. Pipeline is best for sequential, stage-by-stage workflows. Swarm works for exploratory or creative tasks where emergent consensus beats top-down planning. Peer-to-peer mesh suits research teams that need rich cross-agent context sharing. See the [comparison of hierarchical vs peer-to-peer agents](/blog/hierarchical-vs-peer-to-peer-agents/).

When should I use a single agent instead of a multi-agent team?

Use a single agent when your task fits within one context window, requires no parallelization, and doesn't need specialized expertise from distinct domains. Multi-agent teams make sense when tasks hit context window limits, when parallel execution would save significant time, or when you need specialist models for different subtasks. Start single, then add agents only when you hit a real limit.

AI Agent Team Composition: Roles, Skills & Topology

COMPLETE guide to AI agent team composition — roles, skill assignment, and communication topologies. Design multi-agent teams that actually work. Start free on cowork.ink.

Frequently Asked Questions

What roles should an AI agent team have?: Every effective AI agent team needs at least an orchestrator (the planner that decomposes tasks), one or more specialist workers (domain-focused executors), and a verifier (critic agent that checks outputs). Larger teams also benefit from a router agent for intent classification and a memory/state agent for shared context. See our [guide to types of AI agents](/blog/types-of-ai-agents/) for definitions of each.
What is the difference between an orchestrator agent and a worker agent?: An orchestrator agent handles high-level planning — it breaks a complex task into subtasks and delegates them to worker agents. Worker agents are specialists that execute one well-defined capability (e.g., web search, code generation, data analysis). Orchestrators synthesize results; workers produce them. The division is analogous to a project manager and individual contributors on a software team.
How many AI agents should be on a team?: Start with 3–5 agents for most tasks. Anthropic's production multi-agent research system runs 3–5 subagents in parallel. Beyond 7–8 agents, coordination overhead typically outweighs the parallelization benefit. The right number depends on task parallelizability, not on "more is better." Add agents only when hitting a concrete operational limit — context window exhaustion, speed, or specialist knowledge gaps.
What communication topology is best for AI agent teams?: It depends on your task structure. Hierarchical (orchestrator-worker) is best for complex tasks requiring central coordination. Pipeline is best for sequential, stage-by-stage workflows. Swarm works for exploratory or creative tasks where emergent consensus beats top-down planning. Peer-to-peer mesh suits research teams that need rich cross-agent context sharing. See the [comparison of hierarchical vs peer-to-peer agents](/blog/hierarchical-vs-peer-to-peer-agents/).
When should I use a single agent instead of a multi-agent team?: Use a single agent when your task fits within one context window, requires no parallelization, and doesn't need specialized expertise from distinct domains. Multi-agent teams make sense when tasks hit context window limits, when parallel execution would save significant time, or when you need specialist models for different subtasks. Start single, then add agents only when you hit a real limit.