coala
hubentu 作
DevOps & Cloud概要
How to use the coala-client CLI for chat with LLMs, MCP servers, and skills.
説明
このスキルでできること
coala-client is a CLI tool for chatting with OpenAI-compatible LLMs and connecting to MCP (Model Context Protocol) servers. It supports importing CWL toolsets as MCP servers and loading skills into chat context. Works with OpenAI, Gemini, and Ollama providers.
Combines LLM chat and MCP tool execution in a single CLI without needing separate clients for each provider or server.
使いどころ
- Chat with a local Ollama model via terminal
- Import a CWL toolset and call its tools from the command line
- Switch between OpenAI and Gemini mid-session
- Load a skill into context with /skill to guide the LLM
- Run a one-off query against MCP tools with coala ask
ワークフロー例
User asks: how do I query gene TP53 data using an MCP toolset?
- 1
Run coala mcp-import gene-variant <URL-to-cwl> to import the toolset
- 2
Run coala mcp-list gene-variant to inspect available tools and their schemas
- 3
Start coala chat to enter interactive mode
- 4
Call the tool directly with coala mcp-call gene-variant.ncbi_datasets_gene --args '{"data": [{"gene": "TP53", "taxon": "human"}]}'
JSON response from the MCP tool with gene data for TP53
要件
- OPENAI_API_KEY environment variable (for OpenAI provider)
- GEMINI_API_KEY environment variable (for Gemini provider)
- coala-client binary installed via uv
インストール
clawhub install hubentu/coalaCLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。