Multi-agent dev experiment fails at integration: 'I became the API'
LuckFantastic7572 · reddit · 2026-10-17
A developer building an options strategy scanner with a small AI agent team — ChatGPT for coordination, two Claude agents for coding and review — recounts how three integration approaches (custom file bridge, Slack triggers, n8n orchestration) all failed to reliably chain agents end-to-end, leaving him manually passing information as the "human API."
The revised architecture:
- Dropped the ChatGPT coordinator; the two Claude agents now collaborate directly
- Switched from polling to event-driven execution, where task completion triggers the next step
- Kept safety boundaries intact — the dev agent correctly refused to merge code due to permission limits, and the human did the merge
Results are too early to call; the author is asking the community for real-world multi-provider agent orchestration setups (API, MCP, n8n, custom event handlers).
More from coding & agent
- Cloudflare open-sources Pipecat voice agent on Python Workers with WebRTC — ritakozlov · 2026-10-17
- LEMO agent pipeline runs on real robots, completing supervised tennis-ball-in-cup task — du_yilun · 2026-10-17
- Having your agent read 8,000-word articles may be the future of serious online writing — danshipper · 2026-10-17
- Tencent open-sources Hy-Memory 2.0 for cross-tool agent memory sharing — aigclink · 2026-10-17
- OpenSEO: open-source SEO platform with MCP lets AI agents do keyword research for $10/month — diegobuildsai · 2026-10-17
- Approval flows should follow the risk, not the org chart — usehive · 2026-10-17