Matrixx is highly inspired by oh-my-opencode — the project that pioneered the "oh-my-zsh for OpenCode" concept. Full credit to code-yeongyu for the original vision.
Multi-model agent orchestration for OpenCode.
14 specialized agents. ~52 lifecycle hooks. 28 tools. One plugin.
OpenCode is a powerful open-source AI coding agent. Matrixx makes it smarter.
Instead of one model doing everything, Matrixx coordinates a team of specialists — each model doing what it does best, in parallel, with full context awareness. The right model for the right job, automatically.
You: "Add OAuth2 with PKCE to the API"
↓
Morpheus (Claude Opus) → Plans the implementation
├─ Keymaker (GPT 5.3) → Builds auth middleware + routes
├─ Oracle (Claude Sonnet 4.6) → Reviews architecture in parallel
└─ Sentinel (Sonnet 4.6) → Audits for security vulnerabilities
↓
Done. Tested. Secure.
| Problem | Matrixx Solution |
|---|---|
| One model does everything poorly | 14 specialists — right model for the right job |
| Agent forgets what it was doing | Todo Continuation — forces completion, no exceptions |
| Slow sequential tool calls | Parallel background agents — 5+ running simultaneously |
| AI-generated code looks like AI | Comment Checker — code indistinguishable from human-written |
| Context window fills up fast | Aggressive delegation — subagents carry the load |
| Fragile refactoring | LSP + AST-Grep — deterministic, safe, surgical |
Don't want to read docs? Just type ultrawork (or ulw) in your prompt.
That's it. Parallel agents, background tasks, deep exploration, relentless execution until completion. The agent figures out the rest.
bunx opencode-matrixx installOr for non-interactive setup (CI/agents):
bunx opencode-matrixx install --no-tui --claude=yes --openai=yes --gemini=no --copilot=nobunx opencode-matrixx doctorCreate matrixx.jsonc in your project root:
Open OpenCode and start coding. Matrixx activates automatically.
For LLM agents — paste this into Claude Code, AmpCode, Cursor, or any LLM agent:
Install and configure matrixx by following the instructions here:
https://raw.githubusercontent.com/klpanagi/opencode-matrixx/refs/heads/dev/docs/guide/installation.md
Installation guide → . Uninstall → . CLI reference →
Matrixx includes a built-in CLI accessible via bunx opencode-matrixx <command>:
| Command | Description |
|---|---|
install |
Interactive setup wizard (or --no-tui for CI/CD) |
doctor |
Environment diagnostics and health checks |
version |
Display version information |
| Category | What It Checks |
|---|---|
| installation | Plugin registration, OpenCode version |
| configuration | Config file validity (matrixx.jsonc) |
| authentication | Provider API key status (Anthropic, OpenAI, Google) |
| dependencies | Runtime deps: Bun, Node.js, Git, Python3 |
| tools | Optional: ast-grep, Gitleaks, PyMuPDF, Playwright |
Use --json for machine-readable output or --category <name> for a specific check.
The one who sees the code for what it truly is.
Role: Master orchestrator and strategic coordinator
Model: Claude Opus 4.6 · temperature: 0.1
Plans, delegates, and executes. Fires background agents in parallel, leverages LSP and AST-Grep for surgical refactoring, and never stops until the TODO list is empty. Morpheus sees the code for what it truly is — and routes every task to the agent best suited for it.
Give him a goal, not a recipe.
Role: Autonomous deep worker
Model: GPT 5.3 Codex · temperature: 0.1
Explores the codebase, matches your patterns, and delivers end-to-end. Keymaker doesn't need step-by-step instructions — give him a destination and he'll find the path, writing production-quality code along the way.
Grammars, parsers, and the art of formal languages.
Role: DSL engineering specialist
Model: Claude Sonnet 4.6 · temperature: 0.1
Grammars, parsers, type systems, code generators, metamodels. 11 composable skills covering textX, ANTLR4, tree-sitter, PyEcore, and more. If it involves defining a language or transforming code, Cipher is your specialist.
Reads every line. Changes nothing. Reports everything.
Role: Read-only security specialist
Model: Claude Sonnet 4.6 · temperature: 0.1
Scans for vulnerabilities but never touches code. OWASP Top 10, SAST, DAST, dependency CVEs, secret detection, crypto audit, infrastructure hardening. 9 composable security skills. Sentinel reports findings with CWE IDs, exact locations, and actionable remediation.
Crafts stunning UI/UX, even without design mockups.
Role: Frontend specialist
Model: Claude Sonnet 4.6 · temperature: 0.1
React/Next.js, Svelte/SvelteKit, accessibility, performance, design tokens, component architecture, build tooling. Sati ships production-grade UI work with browser verification via Playwright. Invoke directly with @sati/ or task(subagent_type="sati") for any non-trivial frontend task.
Architecture demands precision. Oracle delivers it.
Role: Strategic planning, architecture decisions, work plan generation
Model: Claude Sonnet 4.6 · temperature: 0.1
Creates detailed, structured work plans from complex requests. Decomposes ambiguous requirements into atomic, verifiable steps with clear success criteria. Oracle builds the plan — Morpheus executes it.
High-IQ reasoning for problems that refuse to yield.
Role: High-IQ consultation, hard debugging, architecture design
Model: Claude Sonnet 4.6 · temperature: 0.1
Read-only consultation for hard debugging (after 2+ failed attempts), multi-system tradeoffs, and architecture decisions requiring deep reasoning. Merovingian analyzes — never implements.
Where plans become reality.
Role: Plan execution orchestrator, session coordination
Model: Claude Sonnet 4.6 · temperature: 0.1
Executes Oracle's work plans, coordinates session state, manages task dependencies, and ensures every phase completes before moving to the next. The Architect is the bridge between planning and shipping.
Sees what others miss before work begins.
Role: Pre-planning analysis, ambiguity detection, AI failure prevention
Model: Claude Opus 4.6 · temperature: 0.3
Analyzes requests to identify hidden intentions, ambiguities, scope creep, and AI failure points. Seraph intervenes before planning starts — preventing costly mistakes downstream.
Every plan meets Smith's standards — or gets rewritten.
Role: Plan validation, completeness review, gap detection
Model: Claude Sonnet 4.6 · temperature: 0.1
Evaluates work plans against rigorous clarity, verifiability, and completeness standards. Catches gaps, ambiguities, and missing context before implementation begins. Smith is the last line of defense.
Finds what you need, where it lives.
Role: External documentation, OSS search, library research
Model: Claude Haiku 4.5 · temperature: 0.1
Specialized codebase understanding agent for multi-repository analysis, searching remote codebases, retrieving official documentation, and finding implementation examples using GitHub CLI, Context7, and Web Search.
Finds anything, anywhere, instantly.
Role: Blazing fast codebase grep, pattern discovery
Model: Claude Haiku 4.5 · temperature: 0.1
Contextual grep for codebases. Answers "Where is X?", "Which file has Y?", "Find the code that does Z". Fires multiple in parallel for broad searches. Quick, medium, or very thorough — you choose.
Sees what's inside — images, PDFs, diagrams.
Role: PDF, image & diagram analysis
Model: Claude Sonnet 4.6 · temperature: 0.1
Analyzes media files that require interpretation beyond raw text. Extracts specific information or summaries from documents, describes visual content. Use when you need analyzed/extracted data rather than literal file contents.
Small, fast, and disposable — the hands that do the work.
Role: Category-spawned delegated executor
Model: Claude Sonnet 4.6 · temperature: 0.1
Mouse is the worker layer in Matrixx's 3-tier architecture. Spawned automatically when you
use task(category="..."), Mouse executes the task directly without delegating further.
It cannot spawn sub-agents (task tool blocked) — implementation is always done in-house.
Model-specific prompt variants optimize behavior for Claude, GPT, DeepSeek, Mimo, and Qwen.
Every agent, model, temperature, and permission is fully customizable. Meet the full team →
| Agent Orchestration | |
| Developer Tools | LSP (goto def, rename, diagnostics), AST-Grep (search & replace), Tmux terminal |
| ~52 Lifecycle Hooks | Context injection, think mode, comment checking, todo enforcement, error recovery, quality gate |
| 33 Built-in Skills | |
| Curated MCPs | Exa (web search), Context7 (official docs), Grep.app (GitHub code search), Document Reader |
| Claude Code Compat | Full compatibility — commands, agents, skills, MCPs, hooks from settings.json |
| Software Dev Pipeline | 6-phase TDD workflow (PLAN→BUILD→VERIFY→REVIEW→SECURE→SHIP), 5 team roles, adaptive phases |
| Saturation Research | |
| AI Slop Detection |
Full feature list → · Configuration guide → · Architecture diagram →
Matrixx includes a structured 6-phase development pipeline that coordinates specialized roles through PLAN → BUILD → VERIFY → REVIEW → SECURE → SHIP. Each phase has clear entry/exit criteria and is enforced by dedicated agents.
| Role | Agent | Skills | Purpose |
|---|---|---|---|
| Architect | Oracle (Claude Opus) | — | System design, architecture decisions, task breakdown |
| Developer | Source category | git-master, tdd-enforcer |
Implementation code with TDD |
| Tester | Source category | tdd-enforcer, quality-gate |
Test authoring, coverage, verification |
| Quality Evaluator | Red-pill category | quality-gate, review-work |
Lint, typecheck, 5-agent code review |
| Security Expert | Sentinel (Claude Opus) | security-core, security-sast, security-api, security-dependencies |
Vulnerability scanning, CVE checks |
| Phase | Skip? | Role | Exit Criteria |
|---|---|---|---|
| PLAN | Small tasks | Architect | Approach defined, files listed, edge cases documented |
| BUILD | Never | Developer | TDD (RED→GREEN→REFACTOR), bun test passes |
| VERIFY | Never | Quality | lint + typecheck + test + build — all pass |
| REVIEW | Small tasks | Quality (5-agent) | All reviewers PASS, no CRITICAL/MAJOR issues |
| SECURE | Small + non-security | Security | No CRITICAL/HIGH findings, dependencies checked |
| SHIP | Never | Developer | Atomic commits, PR to dev, CI passes |
| Size | Files | Phases Used |
|---|---|---|
| Small | 1-2 | BUILD → VERIFY → SHIP |
| Medium | 3-10 | PLAN → BUILD → VERIFY → REVIEW → SHIP |
| Large | 10+ | ALL 6 PHASES |
| Security-related | Any | Always includes SECURE |
Load the software-dev skill to activate the pipeline. The orchestrator automatically selects the right roles and phases based on task scope.
Matrixx includes a three-tier security layer: reactive hooks, configurable policies, and a dedicated security auditing agent.
Built-in hooks protect against accidental secret exposure — no setup required.
| Hook | What it does |
|---|---|
| Secret Leak Guard | Intercepts git commit and git push, runs gitleaks on staged changes, and blocks the operation if secrets are detected. |
| Env File Write Guard | Blocks agents from writing to sensitive files (.env, *.pem, *.key, credentials.json, id_rsa, and 16 other patterns). |
Both hooks are enabled by default and run before all other hooks in the execution pipeline. Configure via matrixx.jsonc:
{
"security": {
"secret_scanning": { "enabled": true, "block_on_detection": true },
"env_file_guard": { "enabled": true, "allowed_paths": [".env.example"] }
}
}Note: Secret scanning requires gitleaks installed in your PATH. Without it, the hook silently degrades.
Sentinel is a read-only security specialist with 9 composable skills covering the full application security stack:
| Skill | Domain |
|---|---|
security-core |
OWASP Top 10, CWE classification, threat modeling (STRIDE) |
security-secrets |
Secret detection, credential scanning, pre-commit hooks |
security-sast |
Static analysis, code vulnerability patterns, taint tracking |
security-dast |
Dynamic analysis, runtime testing, fuzzing, penetration testing |
security-dependencies |
CVE scanning, SBOM generation, supply chain security |
security-api |
Authentication, authorization, CORS/CSRF, input validation |
security-crypto |
Encryption audit, key management, TLS, password hashing |
security-infra |
Container scanning, Dockerfile hardening, IaC audit, K8s security |
security-review |
Structured audit reports, severity classification, remediation guidance |
Sentinel never modifies code — it reports findings with CWE IDs, exact locations, and actionable remediation. Any agent can load individual security skills via load_skills.
Matrixx integrates RTK for automatic bash command compression, reducing LLM token consumption by 60-90% on tool outputs.
RTK is a Rust CLI binary that intelligently rewrites bash commands to compress their output before it reaches the LLM. It recognizes 70+ command patterns (git, npm, cargo, test runners, linters, build tools) and applies smart filtering, grouping, and deduplication strategies.
# Without RTK: 2000 tokens
$ git status
On branch main
Changes not staged for commit:
modified: src/config.ts
modified: src/hooks/index.ts
... (50 more lines)
# With RTK: 200 tokens
$ rtk git status
2 files changed: src/config.ts, src/hooks/index.ts
The RTK hook intercepts bash commands before execution and rewrites them to use RTK's compression:
- LLM requests:
git status - RTK hook rewrites to:
rtk git status - RTK binary executes and compresses output
- Compressed output (60-90% smaller) reaches the LLM
The hook runs silently — no configuration needed beyond enabling it. RTK's pattern matching handles the rest.
RTK is disabled by default (opt-in). Enable it in matrixx.jsonc:
{
"rtk": {
"enabled": true,
"binary_path": "/usr/local/bin/rtk", // optional — defaults to "rtk" in PATH
"timeout_ms": 5000 // optional — subprocess timeout
}
}Install RTK from rtk-ai/rtk:
# macOS
brew install rtk-ai/tap/rtk
# Linux (curl)
curl -fsSL https://raw.githubusercontent.com/rtk-ai/rtk/main/install.sh | bash
# Verify installation
rtk --version| Metric | Value |
|---|---|
| Overhead | ~10-20ms per bash command |
| Token savings | 60-90% reduction on compressed commands |
| Net benefit | Significant for projects with frequent bash commands |
The 10-20ms subprocess overhead is negligible compared to command execution time and LLM context savings.
| Overview | What Matrixx does, workflows, getting started |
| Agents Deep Dive | Full agent descriptions, skills, workflows, example prompts |
| Architecture | System diagrams, delegation flows, model routing |
| Features | Complete feature reference |
| Configuration | All config options, agent overrides, hooks, categories |
| Orchestration | How agents coordinate, delegate, and recover |
| Categories & Skills | Task categories, skill injection, delegation patterns |
If this saves you time, a ⭐ goes a long way.
Productivity might spike too hard. Don't let your coworker notice. Actually — let's see who wins.
{ "$schema": "https://raw.githubusercontent.com/klpanagi/opencode-matrixx/refs/heads/dev/dist/matrixx.schema.json", "agents": {} }