Scaling to 200+ tools without route failures: Playbooks over model improvisation
styleforge-io · reddit · 2026-08-17
The author shares two years of building a creative OS for agencies, now with 200+ tools and 500+ APIs. They found that model-improvized routes cause compounding errors. Solution: 150+ written playbooks defining steps, order, tools, and done criteria. System organized into three lanes (creative, intelligence, account) to narrow decisions. Two limits: can't fix silent tool failures, and index must be in the path. Asks community for solutions to multi-step tool call issues.
More from coding & agent
- Private AI Deployments Pose Greater Risk Than Public Models — maksym_andr · 2026-08-17
- Grok 4.6 Coding Test: Strong Core Logic, Weak Boundary Handling — jquinonero · 2026-08-17
- Grok revealed as a subagent in architecture debate — PawelHuryn · 2026-08-17
- Unofficial Google Health MCP Server: Query Fitbit Data Locally — delxmobile · 2026-08-17
- Orca: Open-source tool to run five coding agents and pick the best result — dr_cintas · 2026-08-17
- Poll: How coding with LLMs shifts or distorts your sense of time — StewartalsopIII · 2026-08-17