Version: 1.7.13

Overview

OpenCode Orchestrator coordinates multi-agent workflows with autonomous verification and local-first memory.

🌐 Interactive 3D Explorer: Experience the complete 3D runtime architecture, agent council, hybrid RAG, and Ebbinghaus memory in your browser: https://agnusdei1207.github.io/opencode-orchestrator/

• Autonomous Mission Loop: Commander, Planner, Worker, and Reviewer collaborate with verification gates before concluding tasks.
• Role-Aware Context: Pluggable agent profiles with modular system prompts and tailored retrieval weights.
• Local-First Memory: On-disk Ebbinghaus decay model combining BM25, tags, and graph connections without external vector databases.
• Fast Rust Tooling: High-performance AST, LSP, search, and parallel execution primitives.

1. Installation

npm install -g opencode-orchestrator

Note: The install hook automatically registers the plugin in opencode.json / opencode.jsonc.

To remove the plugin:

npm explore -g opencode-orchestrator -- npm run cleanup:plugin
npm uninstall -g opencode-orchestrator

2. Configuration

Add or customize in opencode.jsonc:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": [
    [
      "opencode-orchestrator",
      {
        "agentConcurrency": {
          "commander": 1,
          "planner": 10,
          "worker": 10,
          "reviewer": 10
        },
        "missionLoop": {
          "ledger": true,
          "markdownMemory": true
        }
      }
    ]
  ]
}

• Model Inheritance: Subagents inherit the primary agent model unless explicitly configured under agent.<name>.model.
• Options Schema: Full configuration schema is available in opencode-orchestrator.schema.json.

3. Usage

Start a mission inside OpenCode:

/task "Implement feature and verify test suite"

| Command | Action | | --- | --- | | /task <objective> | Starts a persisted mission loop under .opencode/ | | /stop or /cancel | Halts the active mission loop | | Esc (Interrupt) | Pauses loop continuation until next turn |

4. Multi-Agent Architecture

                    /task input
                         |
                         v
                   +-------------+
            +----->|  Commander  | (Coordinates & delegates)
            |      +------+------+
            |             |
            |       +-----+------+
            |       v            v
            |  +---------+  +-------------+
            |  | Planner |  | Worker Pool | (TDD & isolated edits)
            |  +----+----+  +------+------+
            |       |              |
            |       v              v
            |  +---------------------+
            |  |  Mission Workspace  |
            |  |     (.opencode/)    |
            |  +----------+----------+
            |             |
            |             v
            |       +------------+
            |       |  Reviewer  | (Verifies evidence & tests)
            |       +-----+------+
            |             |
            |             v
            |      +-------------+
            +------|  Verified?  |
             (no)  +------+------+
                          | (yes)
                          v
                        Done

| Agent | Role | Responsibility | | --- | --- | --- | | Commander | Orchestrator | Interprets missions, manages agent pool, and handles loop state. | | Planner | Strategy | Breaks objectives into ordered file-level tasks and tracks TODOs. | | Worker | Implementation | Executes file changes with isolated context and TDD practices. | | Reviewer | Gatekeeper | Verifies test evidence, builds, and marks completion. |

5. Shell Listener (Optional)

For authorized testing environments, a multi-session TCP shell listener TUI is available via the bundled Rust CLI:

orchestrator shell-listener --bind 127.0.0.1 --port 4444

6. Development

# TypeScript
npm run build
npx tsc --noEmit
npm test

# Rust
cargo test --workspace
cargo clippy --workspace --all-targets -- -D warnings

7. License

MIT License © agnusdei1207