The ora Claude Code plugin: two research agents, written for Claude 5-class models.
Agent frameworks built for weaker models compensate with many agents, hooks, and prescriptive step-by-step instructions. Modern models don't need that. ora ships only what the model can't already know: the outcome it is after, the bar an answer has to clear, and the environment it works in.
Each agent searches in its own context and hands back only the answer, so a broad search doesn't fill up your main conversation.
exploreanswers questions about the local codebase. It is read-only.researchanswers questions from docs, public repos, package registries and the web.
Neither agent is taught a method. They work through Claude Code's Bash tool with the CLIs you have installed; the plugin README lists the ones they look for.
The CLIs the agents use, which the plugin does not install. The plugin's Brewfile covers the ones in Homebrew, and the plugin README covers the rest:
brew bundle --file ~/.claude/plugins/marketplaces/agentskills/plugins/ora/Brewfile/plugin marketplace add trancong12102/agentskills
/plugin install ora@agentskills| Component | Credential | How to get |
|---|---|---|
research |
TINYFISH_API_KEY |
Create a key at agent.tinyfish.ai/api-keys, then export TINYFISH_API_KEY=…. Parallel page fetches, search and browser agents through the tinyfish CLI. |
| answer audit (optional) | JEV_API_KEY |
Create a key at console.typesafe.ai/keys, then export JEV_API_KEY=… (TYPESAFE_API_KEY also works). Turns on Jev's two extra checks in scripts/audit-answers.py. |
Neither is read at startup. Without TINYFISH_API_KEY, research falls back to the built-in web tools; without JEV_API_KEY, the audit runs its offline check alone.
MIT. Copyright Cong Tran.