Skip to main content
plugin

hivemind — Claude Code plugin: opencode workers and swarms

Hands grunt work to opencode workers running cheap or open models while Claude keeps planning and merging. Code leaves your machine by default.

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add alirezarezvani/claude-skills
/plugin install hivemind@claude-code-skills
Open on GitHub

What hivemind does

The plugin spins up a single opencode worker or a swarm split into scout, coder and tester roles. Claude stays in charge: it plans the task, reviews what comes back and merges the result, while git worktrees keep each worker isolated. It ships outside every domain bundle, so nothing pulls it in automatically.

Think about where your data goes. The default worker model opencode/mimo-v2.5-free runs on opencode's hosted service, so every prompt, file and diff a worker sees is sent to their endpoints. No worker-side call touches the Anthropic API. Pass --model pointing at a locally served model — an Ollama or LM Studio provider set up in your opencode config — and the code stays with you.

You need the third-party opencode CLI and an opencode account. With a hosted model, delegate only code you are allowed to send to that provider, and keep secrets out of worker prompts.

What it helps with

  • cut spend on bulk edits that would burn paid tokens
  • search a repo with a dedicated worker
  • run coding and testing in parallel across three roles
  • isolate experiments in git worktrees
  • switch a worker to a local model

Example prompts for the agent

  • Start an opencode worker on opencode/mimo-v2.5-free and have it find every call to the old API
  • Set up a scout/coder/tester swarm: scout maps the edit points, coder writes the patch, tester runs the suite in its own worktree
  • Rerun the worker with --model pointed at a local model from my opencode config so the code stays here

How to install hivemind

Plugin marketplace

/plugin marketplace add alirezarezvani/claude-skills
/plugin install hivemind@claude-code-skills

Step by step, with commands for each platform: Claude Code plugins: marketplaces, installing, and building your own

Questions about hivemind

How do I install hivemind in Claude Code?

Add the marketplace with /plugin marketplace add alirezarezvani/claude-skills, then install the plugin: /plugin install hivemind@claude-code-skills.

What does hivemind do?

It lets Claude orchestrate opencode workers — one at a time or a scout/coder/tester swarm — on cheaper and open models. Claude plans, reviews and merges.

Where does my code and prompts go?

The default model runs on opencode's hosted service, and everything a worker receives is sent there. Point --model at a local model and the code stays on your machine.

Source text

hivemind · MIT
# hivemind

Opt-in plugin that lets Claude Code orchestrate [opencode](https://opencode.ai) workers
(single worker or a scout/coder/tester swarm) on cheaper or open models. It ships
outside every domain bundle; install it only if you want it.

## Data flow

**By default your code leaves your machine.** The default worker model,
`opencode/mimo-v2.5-free`, runs on opencode's hosted service, so every prompt, file and
diff a worker receives is sent to opencode's endpoints. Nothing on the worker side
calls the Anthropic API.

To keep code on your machine, pass `--model` pointing at a locally served model (for
example an Ollama or LM Studio provider configured in your opencode config). With any
hosted model, delegate only code you are allowed to send to that provider, and never
delegate secrets.

## Install and use

See [SKILL.md](SKILL.md) for prerequisites, setup, the invocation contract and the
swarm flow. License: MIT (see [LICENSE.txt](LICENSE.txt)).

© Alireza Rezvani · MIT · alirezarezvani/claude-skills

Similar plugins

All plugins for Agent setup: memory, context, orchestration →

From the same set463

The whole set →
skill-doctor Grades your agent setup against recent local Claude Code and Codex sessions in a repo. Generates a self-contained HTML report with verified skill diffs. ★ 28,733repo stars memory-engineering Audit an agent’s memory costs, records, and forgetting rules. Use it to compare storage designs and catch policies that keep everything indefinitely. ★ 28,733repo stars agent-harness Turns a domain folder of skills into a bounded loop: the agent picks up a goal, compiles it into tasks, executes, verifies and closes only when everything checks out. ★ 28,733repo stars zero-hallucination-coder Enforces a verified workflow before code generation starts, blocking imagined imports and placeholder functions. Intended for multi-file systems, not small edits. ★ 28,733repo stars caveman A compressed communication mode for Claude. The agent drops filler and pleasantries, keeps the technical content intact. ★ 28,733repo stars workflow-builder Designs and writes deterministic multi-agent JavaScript workflows for Claude Code's Workflow tool. It splits complex tasks across fresh sub-agents to save context. ★ 28,733repo stars book-to-skill Builds an agent skill from a book or documentation folder, with chapter files and reference material. Useful when an agent needs to apply a source’s ideas during project work. ★ 28,733repo stars write-a-skill For people writing their own agent skills: walks a new skill through three phases and checks it with bundled Python scripts. ★ 28,733repo stars

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor