Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/ruvnet/ruflo/llms.txt

Use this file to discover all available pages before exploring further.

Ruflo uses environment variables for two purposes: keeping secrets (API keys) out of committed config files, and tuning runtime behavior without changing claude-flow.config.json. Most are optional — only the API key for your chosen provider is required.

​
Example .env File

Copy this as .env.example in your project root, then duplicate it to .env and fill in your actual values. 
# AI Provider Keys
ANTHROPIC_API_KEY=sk-ant-...
OPENAI_API_KEY=sk-...
GOOGLE_API_KEY=...

# Memory
CLAUDE_FLOW_DB_PATH=./.swarm/memory.db
CLAUDE_FLOW_MEMORY_PATH=./.swarm

# MCP
CLAUDE_FLOW_MCP_PORT=3000
CLAUDE_FLOW_TOOL_GROUPS=coordination,memory,monitoring

# Performance
CLI_CORE=0  # Set to 1 for fast lite path (~1.5s vs ~35s cold-cache)
Never commit your .env file to version control. Add it to .gitignore immediately after creating it. Commit only .env.example with placeholder values so teammates know which variables they need to fill in.

​
AI Provider Keys

​
ANTHROPIC_API_KEY
string
required
Anthropic Claude API key. Required when providers.default or providers.fallback is set to claude. Obtain yours at console.anthropic.com.
ANTHROPIC_API_KEY=sk-ant-api03-...
​
OPENAI_API_KEY
string
OpenAI API key. Required for GPT models and Codex. Also used by the OpenAI embedding provider (text-embedding-3-small/large, ada-002).
OPENAI_API_KEY=sk-proj-...
​
GOOGLE_API_KEY
string
Google AI API key. Required for Gemini models (Gemini 2.0 Flash, Gemini 1.5 Pro) and Google-hosted embeddings.
GOOGLE_API_KEY=AIzaSy...

​
Memory Configuration

​
CLAUDE_FLOW_DB_PATH
string
Override the path to the SQLite memory database file. Ruflo resolves the final path using this precedence chain:
  1. --path CLI flag
  2. CLAUDE_FLOW_DB_PATH
  3. CLAUDE_FLOW_MEMORY_PATH/memory.db
  4. .swarm/memory.db in the current working directory
CLAUDE_FLOW_DB_PATH=/var/lib/ruflo/memory.db
​
CLAUDE_FLOW_MEMORY_PATH
string
Sets the directory where Ruflo stores all memory-related files: memory.db, HNSW index snapshots, and RVF exports. Defaults to .swarm/ in the current working directory.
CLAUDE_FLOW_MEMORY_PATH=/home/user/my-project/.swarm

​
MCP Server

​
CLAUDE_FLOW_MCP_PORT
number
Port for the MCP HTTP transport. Defaults to 3000. Ignored when transport is stdio (which is the default for Claude Code integration).
CLAUDE_FLOW_MCP_PORT=3000
​
CLAUDE_FLOW_TOOL_GROUPS
string
Comma-separated list of MCP tool groups to enable. Restricting this list reduces startup latency and the number of tool definitions sent to the model on each turn, which lowers token usage.Available groups: coordination, memory, monitoring, github, system
# Development workflow — coordination + memory only
CLAUDE_FLOW_TOOL_GROUPS=coordination,memory

# Full surface
CLAUDE_FLOW_TOOL_GROUPS=coordination,memory,monitoring,github,system
Precedence: CLI --tools flag > CLAUDE_FLOW_TOOL_GROUPS > mcp.toolGroups in config > all groups
​
CLAUDE_FLOW_TOKEN
string
Internal authentication token used by Ruflo’s security layer. When set, inter-process and HTTP requests are validated against this token. Optional for local development; recommended for any network-accessible deployment. Configure MCP HTTP token authentication programmatically via the auth.tokens array in @claude-flow/mcp server options.
CLAUDE_FLOW_TOKEN=your-secret-token-here

​
Performance

​
CLI_CORE
string
Set to 1 to use @claude-flow/cli-core instead of the full @claude-flow/cli. This is a memory-commands-only lite path with no SQLite, no HNSW index, and no ONNX runtime.Cold-cache npx wall-time drops from ~35s to ~1.5s — a measured 22.9× speedup. Designed for plugin scripts that only need memory_store/memory_search without the full neural stack.
ValuePackageCold-cache timeAvailable features
0 (default)@claude-flow/cli~35sAll features
1@claude-flow/cli-core~1.5sMemory commands only
CLI_CORE=1 npx ruflo@latest memory store key "value"
Eight built-in plugin scripts in the Ruflo monorepo are already CLI_CORE-aware. The full MCP server always uses the complete CLI regardless of this flag. See v3/@claude-flow/cli-core/MIGRATION.md for migration details.

​
Storage

​
IPFS_*
string
IPFS / decentralized storage configuration variables used by the pattern marketplace (transfer-store commands) and the plugin IPFS marketplace. Consult the IPFS documentation for your node’s specific variables.
​
GOOGLE_CLOUD_*
string
Google Cloud Storage configuration variables (project ID, credentials file, bucket name) used by the cloud deployment integration. Used with npx ruflo@latest deployment deploy --cloud gcs.

​
Full Quick-Reference Table

VariableDefaultDescription
ANTHROPIC_API_KEY—Anthropic Claude API key (required for Claude models)
OPENAI_API_KEY—OpenAI API key (GPT models, Codex, OpenAI embeddings)
GOOGLE_API_KEY—Google AI API key (Gemini models)
CLAUDE_FLOW_DB_PATH<cwd>/.swarm/memory.dbSQLite memory database path
CLAUDE_FLOW_MEMORY_PATH<cwd>/.swarmMemory directory (DB + HNSW index)
CLAUDE_FLOW_MCP_PORT3000MCP HTTP server port
CLAUDE_FLOW_TOOL_GROUPS(all groups)Comma-separated list of enabled tool groups
CLAUDE_FLOW_TOKEN—Internal authentication token
CLI_CORE0Set to 1 for fast lite path (22.9× faster, memory-only)
IPFS_*—IPFS / decentralized storage config
GOOGLE_CLOUD_*—Google Cloud Storage config
In production, set CLAUDE_FLOW_TOOL_GROUPS to the minimum set your workflow requires. Restricting to coordination,memory instead of all five groups shaves tool-list tokens on every model turn and reduces the attack surface of your MCP endpoint. Add github only in repos where agents are explicitly authorized to open PRs or triage issues.

Build docs developers (and LLMs) love

Get started for freeTalk to us