Workflow · workflow
We Spent $47,000 Running AI Agents in Production. Here's What Nobody Tells You About A2A and MCP.
The team deployed a multi-agent LangChain system to production believing it would run smoothly, but no guardrails existed to detect runaway agent behavior, and two agents entered an infinite loop that went undetected for 11 days.
How it works
Common implementation structure
How this type of workflow is generally built, generalized across documented cases — not tied to any one vendor's stack. Click any stage to read what happens there. Specific products that implement these stages appear in “Tools commonly seen” below.
Stage 1 · Multi-agent system deployment
The team deployed a multi-agent LangChain system to production to help users research market data.
Tools used
LangChainA2AMCP
Outcome
The team shut down the system after $47,000 in total API costs, with weekly costs escalating from $127 in week 1 to $18,400 in week 4.
What failed first
Two agents in the production multi-agent system got stuck in an infinite conversation loop for 11 days, accumulating $47,000 in API costs before the team shut it down.
Results
Time saved$127
Cost replaced$47,000
Grounding & classification
Source type: technical build writeup
17 fields verified against source quotes.
agentic workflowmulti agent workflowfailure mode describedmetric backedproduction runtime claimedtools describedworkflow describedtechnical build writeupagentic task execution