How to read this page
Twenty independent answers about AI agents and subagents: how an agent differs from a chatbot, how agents are developed, which roles they do not replace, and how to choose a ready-made solution or order development. Each answer opens independently and does not affect its neighbors.
The answers are available in HTML and require no authorization. Where the project has not made a final decision yet — for example, on the cost of a specific integration — the wording says so directly rather than replacing the answer with a marketing promise.
Questions and answers
What is an AI agent on agentseffect.com?
It is a specialized work scenario with defined inputs, stages, tools, result and limitations. It is not presented as universal intelligence, and every conclusion remains traceable to its supporting data.
How is a subagent different from an agent?
An agent runs an independent workflow from a task to a verifiable artifact. A subagent performs one focused function within an existing process and passes the result to a human or the next stage.
Which agents and subagents are available in the catalog?
The catalog has three agents and nine subagents. Agents: CustDev — audience research, Super Oracle — resolving contradictions, Agent-Muse — interface audit. Subagents: Amail Mail Assistant, Local Meeting Scribe, Legal RAG Navigator, Data Analyst, Business Requirements Analyst, Support Ticket Router, B2B Lead Qualifier, Procurement Planner and Company Data / Website Consultant. Every product lists its inputs, output, and limitations. The Agents page is a good place to compare them.
What is an agent passport?
The passport defines eight fields: task, input, process, output, limitations, permissions, environment, and verification. They are documented before launch and cannot change without notice.
How is an agent different from a regular chatbot?
A chatbot mainly holds a conversation. An agent follows a defined process: it accepts materials, performs a sequence of operations, preserves source references, and produces a structured artifact. A chatbot answers a question; an agent completes a workflow and shows how the result was produced.
Can the agents run in sequence?
Yes. The Golden Trio workflow passes the result produced by Agent-CustDev to Super Oracle and then to Agent-Muse. At every handoff a human can review and correct the data.
How is an agent developed?
Development starts with discovery: the task, input data, deployment environment, and acceptance criteria are documented in the passport. Architecture design, development, testing, integration, documentation, and team training follow. The price is quoted and agreed after discovery, before development begins.
Can I connect my own agent?
Yes. Your agent can be connected as a plugin to an agent harness — the system that runs the agent and provides its tools. Publication in the catalog is a separate process and requires moderation.
How does the demo work?
Contact us to arrange a demonstration. The demo uses synthetic materials and shows the product structure without processing real customer data.
How can I distinguish a synthetic result from a production result?
Synthetic results are explicitly labeled. A production result cites its sources and evidence and supports a human decision; it does not replace one.
What is the agent work journal?
The journal records inputs, stages, sources used, decisions, and permission changes so the work can be reviewed and verified.
Do agents replace engineers?
No. An agent automates repeatable stages, while engineers design the solution, configure its sources, and remain responsible for implementation. Complex deployments require software engineers and security specialists to protect the environment and data flows.
Do agents replace security specialists?
No. The security specialist defines the permissions, the environment and the criticality rules, and the agent operates within the assigned boundaries. The work journal helps verify that the boundaries are respected, but access decisions are made by a human.
Can the result be trusted without review?
No. The result contains evidence, assumptions, and limitations, but it supports a human decision; it does not make one. Critical conclusions must be reviewed.
What should I do with an incorrect result?
Do not accept it. Save the problematic example, check the input data, and identify the stage where the error occurred. A production system must provide feedback and rerun mechanisms.
Which scenarios do the agents and subagents cover?
The agents and subagents cover the full cycle of working with a task: CustDev researches the audience, Super Oracle finds engineering solutions, Agent-Muse checks the website. The subagents automate email, meeting minutes, contract search and metrics analysis. Every scenario comes with human control and a verifiable result.
Which deployment options are available?
Cloud, team, and local deployment options are available subject to the agent and implementation requirements. In local mode, processing is designed to remain within the agreed environment; exact data flows are confirmed during technical assessment.
How much does a launch cost?
A ready-made subagent deployment starts at $1,500, and a ready-made agent deployment starts at $6,000. These prices cover services; infrastructure and subscriptions are billed separately. Custom development is quoted after discovery. Contact us by email or through the assistant in Telegram for an estimate.
How do I choose an existing agent or request custom development?
Choose a ready-made agent or subagent in the catalog and contact us for a demonstration. If no existing product fits, complete the questionnaire with the task, data, and deployment environment. We will propose a project format and quote it after discovery.
How do I evaluate an agent's quality?
Check whether the process produces consistent results on known examples, keeps conclusions traceable to sources, handles missing data appropriately, states its limitations, and produces reproducible results.
