Most developers hit the same limit: context runs out, the AI forgets your project, you re-explain before asking real questions. Claude Orchestrator Pro eliminates that ceiling.
It's not a VS Code extension. It's a workflow system that turns your editor into an AI command center with persistent memory, coordinated agents, and safety rails.
A Setup That Configures Itself
Orchestrator Pro includes a cross-platform install.py that handles the entire setup. It detects your environment (GPU/CPU/Apple Silicon), downloads local models, starts containers, and registers agents and skills.
Zero to operational: under fifteen minutes, once.
The VCT Launcher is your desktop hub for launching and updating afterwards.
A Knowledge Graph That Remembers Everything
The Orchestrator's local vector knowledge graph (Weaviate + Ollama's qwen3-embedding) and separate code graph (CodeSage-Large-v2) both run on your machine. Always included, no paywall on memory.
Every decision, pattern, and conversation gets indexed and retrieved in future sessions. The AI searches your actual project history instead of starting blank.
Pro adds RL-scored retrieval reranking — a neural network that learns from your clicks and surfaces the results you actually use. Pro also adds curated agent packs and auto-updates.
The AI in week twelve has context from week one. No re-explaining.
Search by project, concept, or tool — get real answers, not hallucinations.
16 Base Agents (Plus 10 in MAO)
The base Orchestrator includes 19 specialized agents (free, AGPL-3.0) — each with defined role, system prompt, and scope. MultiAgent Orchestrator (MAO) adds 10 more, totaling 26 when installed. Pro adds no new agents, just RL reranking, curated packs, and auto-updates.
Examples:
- Planner: decomposes goals into phased plans
- Code Reviewer: audits diffs for bugs, style, security
- Tester: writes unit and integration tests
- Documenter: generates inline docs from code
- Debugger: proposes fixes from stack traces and logs
Complex tasks run parallel workflows — Planner hands off to Coder while Tester scaffolds tests. No single chain bottleneck, coordinated team execution.
29 Skills
The base Orchestrator includes 28 slash-command skills — discrete, invokable units composable into workflows. All 29 in free release; Pro doesn't change this.
Skills cover:
- Semantic search across KG
- Code graph analysis (callers, dependencies)
- Web research with auto-capture
- Brainstorming and spec generation
- Context compaction
- PDF and document ingestion
Agents call skills mid-task, or invoke directly with /skill-name.
16 Automation Hooks
AI with file system access needs guardrails. The Orchestrator runs 20 automation hooks around every tool call, write, and edit. All 16 in free release.
Hooks handle:
- Credential scanning (blocks AWS keys, API tokens)
- KG auto-sync on edits
- Code graph updates on file changes
- Context snapshots and re-injection
- Bash safety rules
- Environment scrubbing before subprocesses
Out of the box, they prevent common agentic failures. Customizable.
Four MCP Servers
Orchestrator runs four local MCP servers: Weaviate (knowledge + code graph), Ollama (LLM + embeddings), search (web + code + papers), code-embedding (CodeSage). All four in base release, all local, no third-party for your project memory.
Pricing
VibeCoded Orchestrator is free and open source (AGPL-3.0): knowledge graph, code graph, 19 agents, 28 skills, 20 hooks, 4 MCP servers.
Orchestrator Pro: €19/month, €149/year, €199 lifetime (100 seats max). Adds RL retrieval reranking, curated agent packs, auto-updates. No new agents, skills, hooks, or servers.
MultiAgent Orchestrator (MAO) is separate, currently beta: maestro, planner, hybrid agent, Tauri UI, 10 specialist agents. Pricing at launch.
Not a SaaS wrapper. Install and control it yourself, works independent of third-party services.
Orchestrator Pro is the flagship VibeCoded tool for serious AI in development workflows.
Install free Orchestrator today, then upgrade to Pro.