strategic-advisor-orchestrator

Synaptic Compass is an advisory orchestration framework designed to add a second, more deliberative reasoning layer to AI coding agents such as Claude Code, Cursor, Gemini CLI, and Cline. It positions itself as a peer review system for AI: while the primary agent generates code, a configurable strategic advisor model analyzes the work for architectural soundness, security vulnerabilities, performance issues, and edge cases, then injects structured feedback back into the primary agent's context. The project describes a stateless, event-driven orchestrator with components including an orchestrator module, context aggregator, reasoning engine, feedback interface, and memory buffer. Feedback is categorized as blocking, warning, suggestion, or insight, and the primary agent can accept, override, or escalate the advisor's input. Configuration is handled through a compass.yaml or compass.json file, with options for advisor model selection, validation layers, feedback mode, security scanning depth, and output format. Documented integration status lists Claude Code, Cursor, Gemini CLI, and Cline as stable, with GitHub Copilot and local LSP servers in beta. The README emphasizes local-first operation, stating that no data leaves the environment unless a cloud advisor endpoint is explicitly configured, and supports local models such as Llama 3 and Mistral Large. Use cases include enterprise architecture review, CI/CD security audits, performance tuning, complex bug diagnosis, and startup prototyping with guardrails. The repository is licensed under the MIT License. The README includes a disclaimer that the tool is advisory only and does not guarantee bug-free or secure software.