Open source · Agent orchestration / 2025–2026
Genesis Forge
Bring specialist skills to the tools you already use.
A portable library of specialist agents, skills, and workflows, with local task routing and MCP integration for AI coding clients.
01Coding client
Claude Code, Gemini CLI, or Cline supplies the working environment. The host client handles model access and authentication.
02Rules / MCP
Gemini CLI integrates through a rules file. MCP-compatible clients discover tools such as skill_search, route_task, and get_skill_registry.
03Local routing
Deterministic routing selects relevant agents and skills locally. Genesis Forge does not require a separate model API key for this lookup.
04Skill library
The README describes 24 specialist agents, 79 skill modules, and 13 workflows. These are reusable instructions and expertise supplied to the host client.
05Client execution
The host client uses the selected instructions in its own workflow. Model behavior, tool permissions, and execution remain responsibilities of that environment.
Problem & context
Coding assistants need different expertise for planning, implementation, debugging, and review. Copying those instructions between clients makes them harder to maintain. Genesis Forge packages specialist guidance into a shared library that existing AI coding environments can discover.
Integration workflow
The documented setup command connects rules and MCP configuration to supported clients; a remove command reverses those integrations. Gemini uses a master rules file, while Claude Code and Cline can discover skills through MCP. Node.js consumers can also import the package's skill-path helpers.
Routing & maintenance
Routing runs locally and selects from a skill registry. The repository documents rebuilding that registry with a Python script after adding skills. Keeping discovery separate from model execution lets the same content serve multiple clients, while making registry freshness and client compatibility explicit maintenance concerns.
Scope & limitations
This is an expertise provider within an existing AI environment. The documented agent and workflow counts describe its library, not independently running services or guaranteed task success. This case study does not claim measured token savings, download totals, or benchmarked orchestration quality.