By karanchawla
Define and execute LLM-powered agentic workflows in VVM DSL, where the LLM serves as the runtime. Generate programs from natural language, statically compile for validation and contract inspection, run with multi-agent coordination, parallel execution, semantic control flow, and state backends like SQLite or Postgres, plus boot, inspect, and registry tools.
Initialize VVM and guide the user
Validate and compile a VVM program
Generate VVM programs from natural language descriptions
Inspect a remote workflow contract and cache metadata
Inspect a VVM run from filesystem, SQLite, or Postgres state
Vibe Virtual Machine is a DSL for agentic workflows where the LLM is the runtime.
VVM gives you:
@agent ...)?, match, choose)pmap, parallel(join=..., on_fail=...))input, export, callable modules)in-context, filesystem, sqlite, postgres)The language spec is in skills/vvm/spec.md (v0.0.4).
Install as a Claude Code plugin:
claude plugin marketplace add https://github.com/karanchawla/vvm.git
claude plugin install vvm@vvm
Then:
/vvm-boot
/vvm-compile examples/01-hello-world.vvm
/vvm-run examples/01-hello-world.vvm
For scalable runs:
/vvm-run examples/39-pr-babysitter.vvm --state=filesystem
/vvm-run examples/38-rlm-diff-time-machine.vvm --state=sqlite
input topic: "Topic to research"
agent researcher(model="sonnet", prompt="Research expert.")
agent writer(model="sonnet", prompt="Clear technical writer.")
research = @researcher `Research {topic} and cite evidence.`(topic)
report = @writer `Write an executive summary.`(research)
export report
| Command | Purpose |
|---|---|
/vvm-boot | Initialize and guide first workflow |
/vvm-compile <file.vvm> | Parse/validate without executing |
/vvm-run <file.vvm> | Execute workflow |
/vvm-run-inspect <run-id> | Inspect run state (filesystem/sqlite/postgres) |
/vvm-registry-inspect <source> | Inspect remote workflow contract + cache metadata |
/vvm-generate <description> | Generate VVM from natural language |
See command docs in commands/.
| Mode | Return Type | Best For |
|---|---|---|
in-context | strings | quick local iteration |
filesystem | ref values | long runs, artifact debugging |
sqlite | ref values | local SQL-inspectable state |
postgres | ref values | team observability, high concurrency |
Backends are selected with --state=... on /vvm-run.
VVM supports:
from "./lib/x.vvm" import ...from "https://..." import ...from "@handle/slug" import ...Use:
--offline / --cache-only for cache-only resolution/vvm-registry-inspect to inspect contracts before executionExamples are in examples/, with an up-to-date index in examples/README.md.
Current catalog includes:
01-13, 15-20)21-23, 32-35, 39)24-31)36-37)38)| Path | Purpose |
|---|---|
| skills/vvm/spec.md | Normative language spec |
| skills/vvm/vvm.md | Runtime/execution semantics |
| skills/vvm/patterns.md | Recommended design patterns |
| skills/vvm/antipatterns.md | Common pitfalls |
| commands/ | Slash-command behavior docs |
| examples/ | Runnable reference workflows |
| design/ | Feature-gap and implementation planning docs |
VVM is still evolving. Validate with /vvm-compile before running large workflows, and run high-impact automations with tight permissions and review gates.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub karanchawla/vvm --plugin vvmMulti-agent workflow orchestration via YAML. Ships the conductor skill so the assistant can validate, run, debug, and author workflow files for the conductor CLI.
Orchestrate complex workflows with DAG-based execution, parallel tasks, and run history tracking
Workflow development tools
Multi-agent workflow orchestration with natural language creation, parallel execution, conditional flows, and visual progress tracking
Expert LangGraph/LangChain agent builder with MCP integration, workflow orchestration, and CLI accessibility. Creates production-ready AI agents that are reachable via command line and Claude Code.
This skill should be used when the model's ROLE_TYPE is orchestrator and needs to delegate tasks to specialist sub-agents. Provides scientific delegation framework ensuring world-building context (WHERE, WHAT, WHY) while preserving agent autonomy in implementation decisions (HOW). Use when planning task delegation, structuring sub-agent prompts, or coordinating multi-agent workflows.