Skip to content
forked from ruvnet/ruflo

🌊 The leading agent orchestration platform for Claude. Deploy intelligent multi-agent swarms, coordinate autonomous workflows, and build conversational AI systems. Features enterprise-grade architecture, distributed swarm intelligence, RAG integration, and native Claude Code support via MCP protocol. Ranked #1 in agent-based frameworks.

License

Notifications You must be signed in to change notification settings

blackms/claude-flow

Β 
Β 

Claude Flow

Node >=20 License MIT CI Configured

Mission

Claude Flow is a CLI-first codebase for orchestrating multi-agent workflows and MCP tools in Node.js/TypeScript. The repository currently combines modular V3 packages under v3/@claude-flow/* with legacy integration code under v3/src.

Tech Stack

  • Node.js (declared >=20 at root)
  • TypeScript
  • Vitest
  • npm + pnpm workspace metadata
  • GitHub Actions workflows (ci.yml, v3-ci.yml)
  • Docker Compose assets for RuVector PostgreSQL setup

Features (Implemented Today)

  • CLI command registry with eager and lazy command loading.
  • MCP server management command surface: start, stop, status, health, restart, tools, toggle, exec, logs.
  • MCP stdio mode with JSON-RPC handling (initialize, tools/list, tools/call, ping).
  • RuVector helper commands: setup, import, init, migrate, status, benchmark, optimize, backup.
  • Multi-provider abstraction modules (Anthropic, OpenAI, Google, Cohere, Ollama, RuVector).
  • Plugin SDK packages and plugin implementations under v3/plugins/*.

Architecture

flowchart TD
  A["npm bin: claude-flow"] --> B["v3/@claude-flow/cli/bin/cli.js"]
  B -->|"CLI mode"| C["CLI class"]
  B -->|"MCP stdio mode"| D["MCP JSON-RPC handler"]
  C --> E["Command registry\nv3/@claude-flow/cli/src/commands"]
  E --> F["MCP server manager"]
  E --> G["RuVector commands"]
  E --> H["Providers command"]
Loading
sequenceDiagram
  participant U as User
  participant C as CLI
  participant M as mcp command
  participant S as MCPServerManager

  U->>C: claude-flow mcp start --transport http
  C->>M: resolve subcommand
  M->>S: start(options)
  S-->>M: running status (pid/transport/port)
  M-->>U: startup summary
Loading

Getting Started

Prerequisites

  • Node.js 20+
  • npm

Local CLI Build (required before running bin/cli.js)

cd v3/@claude-flow/cli
npm install
npm run build
node bin/cli.js --help

MCP and RuVector Commands

# From v3/@claude-flow/cli after build
node bin/cli.js mcp status
node bin/cli.js mcp start --transport stdio
node bin/cli.js ruvector setup --output ./ruvector-postgres

Workspace Tests (scripts available in /v3)

cd v3
npm run test:integration:mcp
npm run test:integration:plugin

API Overview

CLI Surface

  • Core orchestration commands include agent, swarm, task, session, memory, workflow, hive-mind.
  • MCP and provider operations are exposed via mcp and providers command groups.
  • RuVector operations are exposed via the ruvector command group.

MCP RPC Methods (stdio mode)

  • initialize
  • tools/list
  • tools/call
  • ping

Roadmap (Future)

  • Remove stale root scripts that reference missing entrypoints.
  • Align documentation claims with automated doc-vs-code checks.
  • Replace scaffolded plugin handlers with fully implemented domain logic.
  • Consolidate legacy v3/src and modular v3/@claude-flow/* boundaries.

Implementation Status

  • Overall status: Partially Implemented.
  • Primary runnable entrypoint is the V3 CLI package build (v3/@claude-flow/cli).
  • QE and Coherence plugin domains expose real interfaces plus scaffold behavior.

Evidence

  • CLI binary mapping: package.json:7, package.json:8
  • CLI runtime behavior: v3/@claude-flow/cli/bin/cli.js:15, v3/@claude-flow/cli/bin/cli.js:146
  • Command registry and lazy loading: v3/@claude-flow/cli/src/commands/index.ts:32, v3/@claude-flow/cli/src/commands/index.ts:63
  • MCP command surface: v3/@claude-flow/cli/src/commands/mcp.ts:747, v3/@claude-flow/cli/src/commands/mcp.ts:750
  • MCP transport options: v3/@claude-flow/cli/src/commands/mcp.ts:70, v3/@claude-flow/cli/src/commands/mcp.ts:75
  • RuVector command surface: v3/@claude-flow/cli/src/commands/ruvector/index.ts:36, v3/@claude-flow/cli/src/commands/ruvector/index.ts:45
  • RuVector setup file generation: v3/@claude-flow/cli/src/commands/ruvector/setup.ts:739, v3/@claude-flow/cli/src/commands/ruvector/setup.ts:750
  • Provider modules and initialization surface: v3/@claude-flow/providers/src/provider-manager.ts:117, v3/@claude-flow/providers/src/provider-manager.ts:129
  • Claude Opus 4.6 provider support: v3/@claude-flow/providers/src/anthropic-provider.ts:61, v3/@claude-flow/providers/src/anthropic-provider.ts:77
  • CI workflows: .github/workflows/ci.yml:3, .github/workflows/v3-ci.yml:6

Known Limitations

  • Root package import entrypoint is deprecated; use CLI (npx claude-flow) or package-level imports (for example @claude-flow/cli).
  • v3/@claude-flow/cli/bin/*.js imports ../dist/src/*; dist is not committed by default.
  • Root workflow compatibility scripts are lightweight wrappers and may no-op when package-level suites are unavailable.
  • Some command outputs are demonstrative rather than telemetry-backed (for example provider usage tables).
  • Performance numbers found in comments/templates are not validated in-repo benchmarks.

Contributing

See CONTRIBUTING.md.

License

MIT, see LICENSE.

About

🌊 The leading agent orchestration platform for Claude. Deploy intelligent multi-agent swarms, coordinate autonomous workflows, and build conversational AI systems. Features enterprise-grade architecture, distributed swarm intelligence, RAG integration, and native Claude Code support via MCP protocol. Ranked #1 in agent-based frameworks.

Resources

License

Code of conduct

Contributing

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages

  • TypeScript 63.4%
  • JavaScript 23.3%
  • Python 9.2%
  • Shell 3.1%
  • Rust 0.4%
  • PLpgSQL 0.4%
  • Other 0.2%