TachikomaAgent Orchestration

Research-backed AI agent system with Plan methodology

Intent Classification

Automatically classifies user requests and routes to optimal strategy based on task complexity.

Structured development with Plan-Apply-Unify loops. Quality over speed-for-speed's-sake.

Context Augmentation & Reinforcement Layer with dynamic rule loading and priority-based enforcement.

Built on peer-reviewed research from AI conferences and arXiv papers.

Dynamic edit format selection optimized for specific LLM models.

Generator-Verifier-Reviser pattern for high-stakes implementations.

Workers for large-context discovery and parallel task execution.

Match task complexity to execution strategy for optimal speed vs accuracy.

Tachikoma can be installed in two locations:

Location	Path	Use Case
Local	`cwd/.opencode`	Project-specific configuration and skills
Global	`~/.config/opencode`	Shared skills and configuration across projects

TIP

Tachikoma automatically discovers skills from both locations, with local installation taking precedence.