These 21 OpenCode Plugins Make AI Development Faster and Smarter

If you spend any serious time writing code in the terminal, you have probably already heard of OpenCode. What began as a lean, open-source coding agent has quietly grown into one of the most customizable AI development tools available — not because of what it ships with out of the box, but because of what its plugin ecosystem lets you bolt on.

This article goes deeper than a simple list. It explains what OpenCode actually is, why its plugin model matters, and walks through a curated set of plugins — drawn from official sources and the rapidly growing community repository — grouped by what they actually help you accomplish.

What Makes OpenCode Different?

OpenCode is a Go-based AI coding agent built for the terminal. It provides a Terminal User Interface (TUI) for interacting with AI models to help you write, debug, refactor, and reason about code. Unlike tools that lock you into a single AI provider, OpenCode is deliberately model-neutral. It connects to over 75 providers including Anthropic Claude, Google Gemini, OpenAI, AWS Bedrock, Groq, Ollama for local models, and many others. You can switch models mid-session, run entirely offline with a local model, or mix and match providers depending on the task.

Under the hood, OpenCode ships with built-in agents — a “Build” agent for active coding and a “Plan” agent for read-only analysis — along with Language Server Protocol (LSP) support, session management backed by SQLite, an auto-compact feature that summarizes long conversations before they hit context limits, and a non-interactive mode useful for scripting and automation. Its plugin system lets you extend nearly all of this without touching the core codebase.

The awesome-opencode community repository on GitHub now lists over 50 plugins, themes, agents, and related projects, with new entries arriving almost daily. What follows is a practical guide to the ones worth your attention.

Memory and Context: Teaching OpenCode to Remember

One of the most common frustrations with AI coding agents is their statelesness. Every new session starts from scratch. You re-explain your project architecture, your preferred patterns, your naming conventions — over and over. Several plugins exist specifically to solve this problem, each taking a different approach.

opencode-supermemory is arguably the most straightforward solution. It hooks into the Supermemory platform to give OpenCode the ability to retain facts, preferences, and context across sessions and across projects. Think of it as the difference between a helpful assistant who forgets everything overnight and one who actually learns how you work over time. For individual developers, the day-to-day productivity difference can be substantial. GitHub: https://github.com/supermemoryai/opencode-supermemory

opencode-agent-memory, inspired by the Letta agent framework, takes a more sophisticated approach. It gives the agent persistent, self-editable memory blocks — meaning the AI can actively update its own memory in response to what it learns during a session, rather than just passively storing what you tell it. GitHub: https://github.com/joshuadavidthomas/opencode-agent-memory

opencode-mem goes even further by using a local vector database to enable semantic memory retrieval. It supports dual memory scopes, a web interface for browsing stored context, an auto-capture system, and multi-provider AI support. If you want truly long-term, searchable knowledge that persists across your entire development history, this is the most feature-complete option. GitHub: https://github.com/tickernelz/opencode-mem

Simple Memory takes the opposite approach — deliberately minimal. It stores memory inside a Git repository, meaning team members can review, commit, and audit what the agent knows. For teams concerned about transparency or working in regulated environments, this low-tech design is actually a feature. GitHub: https://github.com/cnicolov/opencode-plugin-simple-memory

Terminal Power: Going Beyond One-Off Commands

A standard AI coding agent can run a shell command and read back the output. OpenCode-pty goes significantly further than that.

opencode-pty brings true pseudoterminal (PTY) support to OpenCode. This means the agent can launch interactive processes, send follow-up input mid-execution, and return later to read output from long-running operations. Build watchers, interactive REPLs, database shells, and similar tools that would otherwise hang or fail in a basic subprocess model all become manageable. For any developer whose workflow includes long-running background processes, this is a foundational upgrade. GitHub: https://github.com/shekohex/opencode-pty

opencode-background complements this by providing a dedicated background process management layer, letting you queue and coordinate multiple tasks while the agent continues working on the foreground. GitHub: https://github.com/zenobi-us/opencode-background

opencode-background-agents extends the idea further with Claude Code-style async delegation and context persistence — meaning background tasks can carry their own memory and context without interrupting the primary agent’s work. GitHub: https://github.com/kdcokenny/opencode-background-agents

opencode-worktree handles a different but equally practical problem: Git worktrees. It provides zero-friction worktree management that auto-spawns terminals, keeps files in sync, and cleans up on exit. Developers who regularly work across multiple branches simultaneously will find this eliminates a lot of manual overhead. GitHub: https://github.com/kdcokenny/opencode-worktree

The All-in-One Option: Oh My Opencode

If you want a single plugin that upgrades nearly everything at once, oh-my-opencode is the community’s most prominent choice. It bundles background agents, pre-built Language Server Protocol and Abstract Syntax Tree tools, Model Context Protocol integrations, curated agent packs, and a compatibility layer for Claude Code users. The scope is intentionally broad — it is less a single tool and more a comprehensive starting configuration for power users.

For developers who find the full version more than they need, oh-my-opencode-slim is a fork optimized for reduced token consumption. It includes the same specialized sub-agents — Explorer, Oracle, Librarian, Designer, and others — along with background task management, LSP and AST tools, and tmux integration, but is deliberately streamlined to be lighter on context overhead.

GitHub (full): https://github.com/ohmyopencode/oh-my-opencode GitHub (slim): https://github.com/alvinunreal/oh-my-opencode-slim

Model Access and Authentication: Getting More Models for Less

Not everyone wants to pay separately for multiple API subscriptions. Several plugins address this by bridging OpenCode to authentication methods that unlock models through existing accounts.

opencode-antigravity-auth lets you sign in with a Google account via OAuth and access Gemini and Anthropic models through the Antigravity IDE authentication path — without needing a separate API key for each provider. This is especially useful for developers who already have a Google or Gemini plan and want to use it inside OpenCode. GitHub: https://github.com/NoeFabris/opencode-antigravity-auth

opencode-antigravity-multi-auth builds on this with support for multiple Google accounts and automatic rotation when one account hits its rate limit — a practical improvement for heavier workloads. GitHub: https://github.com/theblazehen/opencode-antigravity-multi-auth

opencode-gemini-auth offers a straightforward path to authenticate directly with your existing Google account and activate your current Gemini subscription within OpenCode. GitHub: https://github.com/jenslys/opencode-gemini-auth

opencode-openai-codex-auth takes a similar approach for OpenAI, enabling ChatGPT Plus or Pro subscribers to use the Codex backend directly inside OpenCode via OAuth authentication. GitHub: https://github.com/numman-ali/opencode-openai-codex-auth

Research Workflows: Search with Sources

opencode-websearch-cited adds a web search tool with inline citation support directly inside OpenCode. It can generate source lists alongside search results and supports configurations backed by Google, OpenAI, or OpenRouter depending on your setup. For developers who need to research documentation, API behavior, or recent changes as part of their workflow — rather than switching to a browser and back — this integrates that step directly into the agent session. GitHub: https://github.com/ghoulr/opencode-websearch-cited

The Google AI Search plugin takes a complementary approach by exposing a native tool for querying Google’s AI Mode (SGE) search directly from within OpenCode, useful when you want AI-synthesized answers rather than raw search results. GitHub: https://github.com/IgorWarzocha/Opencode-Google-AI-Search-Plugin

Activity Tracking and Analytics

opencode-wakatime integrates with the WakaTime time-tracking platform to log AI-assisted coding activity, time spent, and file changes. WakaTime’s dashboard gives individuals and teams visibility into how OpenCode is actually being used — which projects, which languages, how much time. For teams that want to measure the impact of AI tools on productivity, or individuals who track their development time for billing or personal review, this plugin feeds that data into an existing, familiar workflow. GitHub: https://github.com/angristan/opencode-wakatime

Tokenscope offers a complementary angle: instead of tracking time, it tracks token usage and cost across providers. Given that token consumption is often the primary cost driver when using cloud AI models, having a per-session breakdown is practically useful for managing expenses. GitHub: https://github.com/ramtinJ95/opencode-tokenscope

Context Analysis provides another dimension — detailed token usage analysis scoped specifically to individual AI sessions, useful for understanding which types of prompts or tasks are heaviest on context. GitHub: https://github.com/IgorWarzocha/Opencode-Context-Analysis-Plugin

Reusable Skills and Workflow Structure

opencode-agent-skills adds support for discovering and loading reusable skills from project folders, user directories, and Claude-compatible locations. Teams can build libraries of prompting patterns, task templates, and domain-specific instructions that any agent can draw from without starting fresh. For organizations standardizing their AI-assisted development practices, portable skill libraries are a meaningful operational asset. GitHub: https://github.com/joshuadavidthomas/opencode-agent-skills

Micode brings structured workflow discipline to OpenCode through a Brainstorm-Plan-Implement cycle with session continuity, subagent orchestration, Git worktree isolation, and AST-aware tooling. It is well-suited for complex, multi-phase tasks where unstructured back-and-forth tends to lose thread. GitHub: https://github.com/vtemian/micode

opencode-handoff solves a specific but real problem: what happens when a long session hits its context limit and you need to continue in a new one? This plugin generates focused handoff prompts that carry the essential context forward, so the next session picks up cleanly rather than starting from nothing. GitHub: https://github.com/joshuadavidthomas/opencode-handoff

Safety, Security, and Sensible Guardrails

As AI agents gain more ability to execute commands and modify files, the case for guardrail plugins grows stronger.

CC Safety Net acts as a pre-execution filter that intercepts destructive Git and filesystem commands before they run. If the agent attempts a git reset --hard, a recursive delete, or similar irreversible operations, the plugin catches it first. The peace of mind alone is worth a few seconds of overhead. GitHub: https://github.com/kenryu42/claude-code-safety-net

Envsitter Guard prevents agents and tools from reading or editing sensitive .env files. It still allows safe inspection — exposing key names and deterministic fingerprints, but never the actual values — so the agent retains enough context to work without ever seeing your credentials. GitHub: https://github.com/boxpositron/envsitter-guard

Dynamic Context Pruning optimizes token usage by removing obsolete tool outputs from the conversation context as a session progresses. Over a long session, tool outputs accumulate and consume context window space even when their contents are no longer relevant. This plugin trims the fat automatically, keeping the context focused and costs lower. GitHub: https://github.com/Tarquinen/opencode-dynamic-context-pruning

Performance and Speed

Morph Fast Apply integrates the Morph Fast Apply API for code editing at over 10,500 tokens per second. It uses lazy edit markers and unified diff output to apply changes with significantly less latency than standard approaches. For developers working on large files or running many rapid iterations, the speed difference is noticeable. GitHub: https://github.com/JRedeker/opencode-morph-fast-apply

opencode-snip automatically prefixes common shell commands — Git, Go, Cargo, npm, Docker, and others — with a filter tool that trims output before it reaches the LLM context window. The project claims 60 to 90 percent reductions in token consumption for affected commands. Over a session with heavy shell usage, that is a meaningful saving. GitHub: https://github.com/VincentHardouin/opencode-snip

Notification Plugins: Know When Work Is Done

If you step away from the terminal while the agent runs a long task, knowing when it finishes matters.

opencode-notify delivers native OS notifications when tasks complete, so you are not staring at a terminal waiting for output. GitHub: https://github.com/kdcokenny/opencode-notify

Smart Voice Notify goes further with voice notifications powered by multiple TTS engines including ElevenLabs, Edge TTS, and SAPI, along with an intelligent reminder system for overdue tasks. GitHub: https://github.com/MasuRii/opencode-smart-voice-notify

Vibe Coding Slack Notifier routes completion alerts to Slack as direct messages, useful for developers who want their task notifications alongside the rest of their team communication. GitHub: https://github.com/Wangmerlyn/vibe-coding-slack-notifier

A Note on Discovery

The plugins above represent a snapshot of a fast-moving ecosystem. The awesome-opencode repository on GitHub (https://github.com/awesome-opencode/awesome-opencode) is the most comprehensive and up-to-date catalogue, currently tracking over 50 plugins alongside themes, custom agents, and related projects. It accepts contributions via pull request, which means the list grows as the community does.

OpenCode itself maintains official SDKs in JavaScript/TypeScript, Go, and Python, making it accessible to developers who want to build their own plugins rather than just use existing ones.