xoonjaeho

Ollama Cc Plugin — AI skill for Claude Code

AI community

Use local/cloud ollama models from Claude Code as a second engine: read-only consult/review, agentic worktree-isolated rescue, and a full Claude Code session driven by an ollama model (/ollama:as-clau.

How to install Ollama Cc Plugin

This entry records only its repository, not the path inside it, so there is no exact command to give. Open xoonjaeho/ollama-cc-plugin and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Ollama Cc Plugin does

Alternatives in AI

Xerj — Local-first search engine and MCP server for AI agents 12.5k ★
Multi Model Consult — Multi-Model Consultation Skill 1.9k ★
Engrim — The Universal Cross-Model Episodic Memory Standard 219 ★

README

ollama plugin for Claude Code

Use local or cloud [ollama](https://ollama.com) models from inside Claude Code as a **second engine** — a quick second opinion, a code review from a different model, or a delegated agentic task, without leaving your workflow.

Two tiers:

Read-only consult/review — a single stateless call to the local ollama daemon (ollama_companion.py).
Agentic rescue — an ollama model runs a tool loop (read/write files, optionally shell) inside a throwaway git worktree and hands you a diff to review (ollama_agent.py). Modeled on the Codex plugin's rescue.

What you get

Command	Tier	Engine	Timeout	Cloud egress	What it does
`/ollama:setup`	—	companion	—	none	check the daemon, list models, flag cloud sign-in
`/ollama:ask`	read-only	companion	5 min · idle-gap	question sent (disclosed)	ask a model a question, answer returned verbatim
`/ollama:review`	read-only	companion	5 min · idle-gap	diff sent (gated)	review your git changes with a model (cloud egress gated)
`/ollama:adversarial-review`	read-only	agent	5 min · idle-gap	reads sent (gated)	a model explores your repo (read/list/grep) and challenges it; steerable focus
`/ollama:rescue`	agentic	agent	30 min · total	reads sent (gated)	delegate a coding task; the model edits files in an isolated worktree → you review the diff before it applies
`/ollama:as-claude`	full session	agent → claude	30 min · total*	reads + actions (disclosed)	run a task in a real Claude Code session powered by an ollama model (`ollama launch claude`) — isolated worktree + diff apply-gate by default, `--no-worktree` / `--resume` write straight to your real tree. Full host shell either way — more dangerous than rescue
`/ollama:list`	read-only	companion	—	none	list installed / available models
`/ollama:ps`	read-only	companion	—	none	list running (in-