The architecture uses an agent credit system as its metering mechanism. Each action type consumes credits at different rates based on computational cost. A single phone call uses approximately 20 credits, reflecting the voice processing and real-time conversation management required. Web searches consume 0.5 credits per query, while email handling uses 2 credits per message. This variable pricing model maps directly to the underlying resource consumption for each task type.
Voice agents support customization across speed, accent, and gender parameters. The system processes natural language input during phone calls and generates appropriate responses based on the agent's defined mission. This works through speech-to-text conversion, language model processing, and text-to-speech synthesis in sequence. Latency depends on call complexity and the number of decision points in the conversation flow.
Email agents parse incoming messages, extract relevant information, and compose responses according to their programmed behavior. They can trigger actions in connected systems based on email content. Web research agents query search engines, extract information from results, and compile findings into structured outputs. The quality of research depends heavily on how well you define the agent's search parameters and evaluation criteria.
Integration happens through connections to email systems, phone infrastructure, and existing business tools. Agent Factory doesn't specify which exact systems it connects to, but agents can coordinate with platforms you're already using. Multi-agent workflows become available on the Force plan, allowing you to chain multiple agents together where one agent's output becomes another's input.
Template-based creation speeds up deployment for common use cases. Templates exist for stand-ups, meeting preparation, appointment scheduling, customer support, sales development, and research tasks. You can modify templates or build agents from scratch if your workflow doesn't match predefined patterns.
The free plan provides 200 agent credits monthly with 1 active agent and 2 seats. That translates to roughly 10 calls, 400 web searches, or 100 emails per month, though you can mix action types. Community support only. The Squad plan at $49 monthly increases capacity to 5,000 credits, 10 active agents, unlimited seats, and standard support. That's approximately 250 calls monthly. The Force plan costs $299 monthly and includes 30,000 credits, 75 active agents, multi-agent workflow capability, and prioritized support. Additional credits cost $99 for 10,000 credits.
Technical limitations center on the agent credit ceiling for each tier. Free users hit the 200-credit limit quickly if running voice-heavy workflows. The 1-agent restriction on free tier prevents parallel task handling. Squad's 10-agent maximum limits organizational scaling. Even Force caps at 75 active agents, which could constrain large operations running numerous specialized agents simultaneously.
The credit consumption model means unpredictable usage patterns can drain your allocation faster than expected. A phone agent that handles complex multi-turn conversations will burn through credits quicker than simple appointment confirmations. Web research agents performing deep investigations consume more credits than basic lookups. You'll need to monitor credit usage closely to avoid hitting limits mid-month. Agent Factory doesn't detail how agents behave when credits run out or whether they fail gracefully or just stop working.
