VITA Engine: Solo Dev Moves AI Agent Orchestration from Code to Markdown Files
Putrid_Bat8818 · reddit · 2026-08-05
A solo developer built VITA, a cognitive engine that shifts agent orchestration from code into plain text (.md files), interpreted by a generic engine at request time.
This approach aims to solve coupling issues in traditional frameworks:
- Decoupled Tool Definitions: Tools are text blocks in Markdown rather than Python functions, removing framework lock-in.
- Dynamic Tool Loading: The model never sees the full catalog, only receiving tools relevant to the current step.
- Hot Reloading: The catalog is reread from disk on every request, allowing capability changes without restarting.
- Provider-Agnostic: Swapping LLM providers requires no changes to tool definitions.
The author notes this isn't meant to replace mature frameworks like LangChain, but rather explores where orchestration should live: in the model, the code, or a declarative layer.
More from coding & agent
- NeurIPS 2026 Call for Papers: Focus on Responsible Meta-Agents — RobertTLange · 2026-08-05
- Spring Evals: Benchmarking LLMs on Real-World Spring Boot 4 Code with Hidden Tests — therealdanvega · 2026-08-05
- Local Coding Benchmark: Qwen3.5-122B Rivals DeepSeek at a Fraction of the Cost — returnity · 2026-08-05
- Navigating SAP's Agent Architecture: MCP for Tool Execution, A2A for Orchestration — SillyDot3305 · 2026-08-05
- YC's Multiplayer Agent Harness: Great Decoupled Architecture, But It Kills KV Cache — Instance_Not_Found · 2026-08-05
- The 'oh-my' Configuration Trend in Open-Source Coding Agents — HarveenChadha · 2026-08-05