Octate
Terminal-native AI code-review CLI combining Tree-sitter AST parsing with LLM critics.
A terminal-native AI code review CLI that combines local Tree-sitter WASM AST parsing, cross-file reference graphs, and static linter discovery with multi-stage LLM review DAGs and a two-stage critic quality gate.
Generic AI code review bots post superficial style comments and hallucinations because they lack whole-repository structural context and AST understanding.
Software engineers and devops leads looking for deterministic, deep code review in terminal and CI workflows.
Parses diffs into Tree-sitter syntax trees locally before prompting models, grounding code analysis in concrete AST nodes and cross-file symbol references.
Verified System Capabilities
Architectural Decisions & Rationales
Local Tree-sitter AST parsing before LLM inference
Rationale: Prevents LLM hallucinations about code syntax, scope, or missing variables.
Two-stage critic quality gate
Rationale: Filters out subjective nitpicks and duplicates before emitting final terminal diff comments.
Technical Trade-offs & Honest Limitations
- Requires language grammar definitions for each supported programming language.
- Large whole-repository cross-file graphs require sufficient local indexing RAM on multi-million line codebases.
Public repository at github.com/eshanized/Octate.