coderica.com

What It Is

hashd is Coderica’s flagship AI development orchestration product. It is not positioned as an IDE autocomplete tool or a standalone code generator; instead, it organizes multiple AI coding agents into an engineering delivery workflow. Starting from REQS.md and the context of the codebase, it handles requirements discovery, planning, micro-commit implementation, automated testing, AI review, human approval gates, PR creation and merging, while preserving full source provenance.

Core Capabilities

On the AI side, hashd supports agents such as Claude Code, OpenAI Codex, Gemini CLI, OpenCode, Kimi Code, and Qwen Code, assigning them to different stages of the workflow. For example, Claude is better suited to architectural analysis, Codex to implementation throughput, Gemini to long-context synthesis, and Qwen to multilingual and documentation-heavy tasks. Its engineering mechanisms include isolated worktrees, parallel workflows, a 13-state workflow engine, three-layer provenance, a real-time TUI dashboard, confidence-based review, and oscillation recovery for detecting agents that repeatedly fail.

Pricing and Integrations

The scraped text does not disclose pricing, free quotas, commercial licensing, or payment methods, so direct costs cannot be assessed. hashd depends on the login and quota systems of external agents, and actual costs will also be affected by each model’s token pricing. On the integration side, GitHub PR creation and merge workflows have been demonstrated. CI/CD integration, RAG knowledge bases, DeepSeek/Grok/local Llama support, metrics dashboards, and other features remain on the roadmap.

Pros and Cons

The main advantage is that hashd moves AI coding from “ad hoc prompts” into an auditable, reversible, and gate-controlled workflow, making it suitable for production engineering teams that care about governance. Local execution without hosting the repository also helps maintain code control. The downside is that if a team only needs quick autocomplete or chat-based coding, hashd may feel heavyweight. If access, permissions, or costs for external agents are difficult to manage, the user experience may also suffer. Some key capabilities, such as RAG, AST-level semantic understanding, and TUI conflict visualization, have not yet been released.

Who It’s For and Access from China

hashd is better suited to teams with established engineering workflows, test systems, and code review requirements, especially organizations that want to centrally manage multiple AI coding tools. The text does not specify access conditions from China. Since the supported agents from Claude, OpenAI, Google, and others typically involve network or payment uncertainty in mainland China, real-world deployment should be verified case by case. Alternative or complementary tools include Cursor, GitHub Copilot, Claude Code, Qwen Code, Kimi Code, OpenCode, and others.