Ecosystem
Community plugins rewritten on the AgentPlugins universal manifest. Each targets functional parity across the four supported harnesses — the same functionality across Claude Code, Codex, OpenCode, and Pi Mono.
Install any plugin with a single command:
bash
agentplugins add owner/repoThis installs the plugin into ~/.agents/plugins/<name>/ and symlinks it to every detected agent on your machine.
Adding your plugin
- Build on the AgentPlugins universal manifest (see Creating Plugins)
- Target all four supported harnesses and confirm functional parity
- Open a PR or issue to add your plugin to this page
See Rewriting for Tier-1 Parity for the step-by-step process.
Available plugins
| Plugin | Description |
|---|---|
| agentplugins-autoresearch | Autonomous experiment loop that gathers what to optimize and runs iterations until the target improves. |
| agentplugins-goal | Autonomous goal-completion loop — set a goal once, the agent iterates until done across all supported harnesses. |
| agentplugins-ponytail | Lazy senior dev mode — forces the minimum solution that works: YAGNI → stdlib → native → one line. |
| agentplugins-caveman | Ultra-compressed communication mode — cuts ~75% of output tokens while preserving full technical accuracy. |
| agentplugins-btw | Parallel side-thread thoughts — spin up side conversations without interrupting the main agent. |
Note: Capability status per plugin lives in each plugin's README. For the harness-level baseline see the Capability Matrix.

