Claude Flow
CautionAI orchestration with hive-mind swarms, neural networks, and 87 MCP tools for enterprise dev.
Install
Claude Code
claude mcp add claude-flow -- npx -y claude-flowSafety Report
details
No built-in authentication. Runs as a local MCP server. Claims CVE-hardened with input validation and command sandboxing via AIDefence.
details
Primarily stdio transport. No HTTP server by default.
details
Local process; no rate limiting.
details
No known CVEs. Self-described as CVE-hardened but no third-party verification available.
details
Large Node.js dependency surface (175+ tools, multiple plugins). No public dependency audit. Caution advised before production use.
details
No dangerous code patterns detected
details
MIT license confirmed.
details
Open source on GitHub (ruvnet/claude-flow). MIT license.
details
Actively maintained. Currently at v3.5 with regular releases.
Compatibility
Tools Provided (35)
swarm_init Initialize agent swarms with configurable topology (hierarchical, mesh, adaptive) and consensus strategy.
agent_spawn Spawn specialized agents of defined types within the swarm.
task_orchestrate Orchestrate task execution across agents with dependency management.
swarm_status Check swarm health and status.
agent_list List active agents in the current swarm.
agent_metrics Retrieve agent performance metrics.
memory_search Semantic vector search across stored agent memory patterns.
memory_store Save patterns with embeddings to the agent memory store.
memory_usage Check memory usage statistics for the agent memory system.
neural_status View the current state of the neural learning system.
neural_train Train the neural system on patterns to improve agent behavior.
neural_patterns List learned patterns in the neural system.
github_swarm Coordinate GitHub operations (PRs, issues) across agent swarms.
repo_analyze Analyze repository structure and codebase.
pr_enhance Enhance pull requests with AI-assisted improvements.
issue_triage Categorize and triage GitHub issues.
code_review Perform AI-assisted code reviews.
task_create Create a new task for agent assignment.
task_list List all tasks and their current states.
task_status Check the status of a specific task.
task_cancel Cancel a running or queued task.
task_assign Assign a task to a specific agent.
task_retry Retry a failed task.
worker_run Execute a background worker process.
worker_status Check the health and status of a background worker.
worker_alerts Retrieve alerts and notifications from workers.
worker_history View execution history of background workers.
ruvector_search Perform vector similarity search using PostgreSQL pgvector.
ruvector_insert Insert vectors into the PostgreSQL vector store.
ruvector_batch_search Perform batch vector similarity searches.
ruvector_health Check PostgreSQL vector store connection health.
progress_check Check V3 implementation progress status.
progress_sync Synchronize progress data across agents.
progress_summary Generate a progress summary report.
progress_watch Monitor progress updates in real-time.
Frequently Asked Questions
What is Claude Flow?
AI orchestration with hive-mind swarms, neural networks, and 87 MCP tools for enterprise dev.
What tools does Claude Flow provide?
Claude Flow provides 35 tools including swarm_init, agent_spawn, task_orchestrate, swarm_status. Each tool can be called by AI agents to perform specific operations.
Is Claude Flow safe to use?
Claude Flow is rated Caution. Some security checks raised warnings. Review the safety report on this page for details before use.
What are alternatives to Claude Flow?
Similar MCP servers include FastMCP, Oh My Posh Validator, PraisonAI. Each serves a similar purpose but may differ in features, language, and compatibility.
Similar MCP Servers
FastMCP
UntestedA high-level framework for building MCP servers in Python
Oh My Posh Validator
UntestedValidate oh-my-posh configurations and segment snippets against the official schema.
PraisonAI
UntestedAI Agents Framework with Self Reflection and MCP support