Sample work · AI & Automation
Agent Design: a multi-agent workflow, verified before construction
For a fictional company: agents with roles, tools, inputs and outputs, handovers, error handling, and cost budgets. A validator identifies built-in design flaws, and a dry run with mocks demonstrates the path to human approval.
Sample work (demo), not a client job. Built as a trial task with invented example data; the check figures below come from the automatic test run and are not polished.
What was built
- Every tool is defined, every handover matches the output and input of the neighbor
- Loops have an upper limit, external impact only occurs via human approval
- Cost budget and error handling per agent; dry run without model calls
- Operations and workflow are fictional
How it was checked
Automatic test run on 2026-10-31, taken from the log (test data invented, figures measured):
- Demo workflow with five agents: 8 built-in design flaws found, after-version 0 findings, dry run with mocks reaches human approval, loop ends at the limit
- 9 checks
Limits
What this sample does not cover, so nobody expects more:
- Checks the design of the workflow, not the quality of the model responses
- Cost budget is a specification, it is only confirmed with measured token counts
- Model or framework (LangGraph, CrewAI, custom code) is only chosen after consultation with the client
Something like this for your business?
Describe briefly what you need. You get a fixed price before I start; nothing happens without your yes.