Keeping multi-agent harnesses alive is a slog: agents fixing agents, and it's sloppy

nptacek · x · 2026-11-20

A developer vents that keeping moderately complex agent harness setups working over time — across bugs, remote session issues, auth failures — is getting really annoying. His usual workaround: ask another agent that's still working to fix the one that isn't, but the whole thing remains sloppy. The post resonates with anyone running long-lived multi-agent workflows.

Original post →

More from coding & agent

coding & agent channel →