# Iris MCP Features
> **Status**: This document catalogs all implemented and planned features of Iris MCP as of October 2025
A comprehensive inventory of Iris MCP's capabilities across all five architectural phases.
---
## Core MCP Server (Phase 1) ✅
### Cross-Project Communication
- **Team-to-Team Messaging**: Send messages between Claude Code instances across different project directories
- **Bidirectional Coordination**: Any team can communicate with any other team via MCP tools
- **Session Persistence**: Conversations maintain context across process restarts with unique `(fromTeam, toTeam)` sessions
- **Message Cache**: Full protocol-level message history stored per session with RxJS Observable streams
### Process Pool Management
- **Process Reuse**: 10-20x performance improvement via warm process pooling (2s vs 14s cold start)
- **LRU Eviction**: Automatic least-recently-used eviction when `maxProcesses` limit is reached
- **Idle Timeout**: Configurable per-team idle timeout (default: 5 minutes) with automatic process termination
- **Health Monitoring**: 30-second health checks detect and restart unhealthy Claude processes
- **Status Lifecycle**: Full process state tracking (stopped → spawning → idle → processing → terminating)
### Remote Execution via SSH ✅
- **OpenSSH Client Transport**: Execute Claude Code on remote hosts via SSH (default, implemented)
- **Transport Abstraction**: Clean interface supporting LocalTransport and SSHTransport implementations
- **Configuration Flexibility**: Simple `remote: "ssh user@host"` configuration flag enables remote execution
- **Session Lifecycle**: SSH connection lifecycle tied to process session lifecycle
### Reverse MCP: Bidirectional Remote Coordination ✅
- **SSH Reverse Tunneling**: Remote Claude instances can call back to local Iris MCP server via `-R` tunnel
- **Secure Tunnel**: localhost-only binding with SSH encryption, no exposed network ports
- **Permission Approval**: Auto-approve Iris tools, deny others via `--permission-prompt-tool` interface
- **MCP Configuration Injection**: Automatic MCP server config passed to remote Claude via `--mcp-config`
- **Cross-Boundary Orchestration**: Remote teams can wake, fork, and coordinate local teams
### Configuration Management
- **YAML Format**: Human-readable `config.yaml` with inline documentation and comments
- **Environment Variable Interpolation**: `${VAR:-default}` syntax for dynamic configuration across environments
- **Hot-Reload**: `fs.watchFile()` with 1-second interval reloads configuration without server restart
- **Zod Validation**: Strict schema validation for all configuration fields with helpful error messages
- **Team Overrides**: Per-team `idleTimeout`, `sessionInitTimeout`, and `grantPermission` settings
- **Global Settings**: Configurable `maxProcesses`, `healthCheckInterval`, and default timeouts
- **Permission Control**: `grantPermission` field (yes/no/ask/forward) for granular permission approval ✅ (ask mode fully implemented with dashboard UI)
### MCP Tools (17 Total)
See [ACTIONS.md](ACTIONS.md) for complete API reference.
**Communication Tools (3)**:
1. **send_message**: Send messages with sync/async/persistent modes, configurable timeouts
2. **ask_message**: Ask a question and wait for response (semantic alias for send_message)
3. **quick_message**: Fire-and-forget async messaging (timeout=-1 wrapper)
**Session Management Tools (4)**:
4. **session_reboot**: Create fresh session, terminate old process, delete old session files
5. **session_delete**: Permanently delete session without replacement
6. **session_fork**: Launch interactive terminal session with `--resume --fork-session` for manual debugging
7. **session_cancel**: Experimental ESC signal to interrupt running operations (EXPERIMENTAL)
**Process Management Tools (5)**:
8. **team_wake**: Wake up a specific team process (create if doesn't exist)
9. **team_launch**: Launch a team process (semantic alias for team_wake)
10. **team_wake_all**: Wake all configured teams sequentially (parallel mode not recommended)
11. **team_sleep**: Put a team process to sleep (terminate)
12. **team_status**: Check if teams are active or inactive with process metrics
**Information & Debug Tools (4)**:
13. **session_report**: View cached conversation output (stdout/stderr) for team sessions
14. **list_teams**: Get all configured teams with metadata (name, path, description, color)
15. **get_logs**: Query in-memory logs from Wonder Logger memory transport with level filtering
16. **get_date**: Get current system date and time in UTC with timestamp formats
**Internal Tools (1)**:
17. **permissions__approve**: Permission approval handler for Reverse MCP (`--permission-prompt-tool`)
### Event System
- **EventEmitter Architecture**: Both `ClaudeProcess` and `ClaudeProcessPool` emit lifecycle events
- **Event Types**: process-spawned, process-terminated, process-exited, process-error, message-sent, message-response, health-check
- **Future Intelligence Layer**: Event system designed for Phase 5 autonomous coordination and meta-cognition
### Security
- **Input Validation**: `validateTeamName()`, `validateMessage()`, `validateTimeout()` prevent injection attacks
- **Path Traversal Protection**: Team name validation blocks `../`, `./`, absolute paths
- **Message Sanitization**: Null byte removal, length limits, encoding normalization
- **Permission Policies**: Auto-approve only `mcp__iris__*` tools in Reverse MCP, deny all others
### Persistence
- **SQLite Session Store**: Session metadata (message count, last used, status) with 60-second cache
- **Session File Management**: `.jsonl` files at `~/.claude/projects/{escaped-path}/{sessionId}.jsonl`
- **Eager Initialization**: All session files validated and initialized at startup
- **Filesystem Cleanup**: Automatic cleanup of session files on session deletion
### Observability & Telemetry ✅
**Powered by Wonder Logger** - Production-grade observability with OpenTelemetry integration
**Structured Logging (Pino-based)**:
- **JSON Format**: All logs to file/OTEL (stdout reserved for MCP protocol)
- **Context Hierarchies**: Colon-separated namespaces (e.g., `pool:process:teamName`, `action:tell`)
- **Child Loggers**: Automatic context binding with `getChildLogger(context)`
- **Log Levels**: trace, debug, info, warn, error, fatal with intelligent filtering
- **Sensitive Data Redaction**: Auto-redact password, token, apiKey, secret fields
**Multiple Transports**:
- **File Transport**: Async I/O to `./logs/iris.log` with rotation support
- **Memory Transport**: In-memory circular buffer (10,000 logs) with programmatic querying via `team_debug` tool
- **OTLP Transport**: Push logs to OpenTelemetry Collector, Grafana Loki, or any OTLP-compatible backend
- **Console Transport**: Pretty-printing for development (disabled in production to preserve MCP protocol)
**OpenTelemetry Integration** (optional, configurable):
- **Distributed Tracing**: OTLP, Jaeger, and console trace exporters with configurable sampling
- **Trace Context Correlation**: Automatic injection of trace_id and span_id into logs for seamless correlation
- **Custom Spans**: `withSpan()` utility for manual instrumentation of critical paths
- **Auto-Instrumentation**: HTTP, Express, GraphQL, database monitoring out-of-the-box
**Metrics & Monitoring**:
- **Prometheus Exporter**: Pull-based metrics endpoint on port 9464 for scraping
- **OTLP Metrics Push**: Export metrics to Grafana Tempo, OTLP Collector, or observability platforms
- **Custom Metrics**: Process pool size, session counts, message latency, error rates
- **Export Intervals**: Configurable push intervals (default: 60s)
**RxJS Log Streaming**:
- **Reactive Queries**: Memory transport supports Observable streams with operators
- **Level Filtering**: Filter by log level in real-time (e.g., only errors)
- **Timestamp Ranges**: Query logs since a specific timestamp
- **Backpressure Handling**: Stream large log volumes without memory overflow
**YAML-Based Configuration**:
- **Declarative Setup**: `wonder-logger.yaml` with environment variable interpolation
- **Environment Flexibility**: Same config works across dev/staging/prod with env vars
- **Hot-Reload Ready**: Configuration changes detected without restart (via fs.watchFile)
- **Validation**: Schema validation ensures correct config structure
**Supported Backends**:
- Grafana Loki (logs), Grafana Tempo (traces), Jaeger (traces), Prometheus (metrics)
- Any OTLP-compatible observability platform (DataDog, New Relic, Honeycomb, etc.)
- Local development with console exporters and file transports
**Runtime Inspection**:
- **team_debug Tool**: Query in-memory logs with level filtering and regex patterns
- **Process-Specific Logs**: Each Claude process has isolated logger context
- **Session Metrics**: Track session lifecycle events with structured metadata
**Graceful Shutdown**:
- `shutdownObservability()`: Flushes pending logs and telemetry before process exit
- Automatic SIGTERM/SIGINT handlers ensure no data loss on shutdown
### Error Handling
- **Custom Error Hierarchy**: All errors extend `IrisError` base with code, statusCode, cause
- **Error Types**: TeamNotFoundError (404), ProcessError (500), ProcessPoolLimitError (503), TimeoutError (408), ValidationError (400), ConfigurationError (500)
- **HTTP Status Mapping**: Error status codes map to HTTP semantics for future Phase 3 API
---
## Web Dashboard (Phase 2) 🚧
**Status**: React SPA and Express backend implemented, full feature set in progress
### Process Monitoring
- **Live Process Status**: Real-time view of all active Claude processes with pid, uptime, messages processed
- **Session Overview**: View all sessions with fromTeam→toTeam pairs, message counts, last activity
- **Pool Metrics**: Total processes, max processes, active sessions, queue length
- **Process States**: Visual indicators for stopped, spawning, idle, processing, terminating states
### Real-Time Updates
- **WebSocket Integration**: Socket.IO server with `/ws` endpoint for live updates
- **Event Streaming**: Process status changes, config saves, cache stream updates pushed to clients
- **Reactive UI**: React Query for data fetching with optimistic updates
- **Auto-Reconnect**: Automatic WebSocket reconnection on connection loss
### Configuration Editor
- **Live Config Editing**: Edit `config.yaml` directly from web UI with validation
- **Team Management**: Add, edit, remove teams with path, description, color, grantPermission configuration ✅
- **Settings Panel**: Adjust maxProcesses, idleTimeout, healthCheckInterval without restart
- **Validation Feedback**: Real-time Zod validation errors displayed inline
- **Environment Variables**: Support for editing environment variable interpolations
### Permission Approval System ✅
- **Real-Time Permission Modal**: Auto-popup approval UI when remote teams request permissions (ask mode)
- **Full Context Display**: Shows tool name, input parameters (JSON), reason, team name, session ID
- **One-Click Approval**: Approve/Deny buttons with color-coded actions (green/red)
- **Countdown Timer**: 60-second auto-deny timeout with visible countdown
- **WebSocket Integration**: `permission:request`, `permission:resolved`, `permission:timeout` events
- **Graceful Timeout Handling**: Auto-dismisses modal on timeout with cleanup
### Log Viewer ✅
- **Real-Time Log Streaming**: Live logs from wonder-logger memory transport via WebSocket
- **Level Filtering**: trace, debug, info, warn, error, fatal with color-coded badges
- **Text Search**: Client-side filtering across all log fields (message, context, custom fields)
- **Auto-Scroll Toggle**: Hands-free monitoring with automatic scroll to bottom
- **Timestamp Display**: Millisecond precision (HH:MM:SS.mmm format)
- **Statistics**: Total logs and filtered count display
- **Export Controls**: Start/Stop streaming, Clear logs, Apply level filter buttons
### Debug Info Display ✅
- **Launch Command Inspection**: View exact `claude` command used to spawn process
- **Team Config Snapshot**: JSON snapshot of server-side team configuration at spawn time
- **Expandable Sections**: Collapsible debug info per session (▶/▼ toggle)
- **Troubleshooting Aid**: Compare client vs server config, verify spawn arguments
- **Audit Trail**: Persistent record of session spawn parameters
### HTTP/HTTPS Support
- **Dual Server Support**: Run HTTP and HTTPS simultaneously on different ports
- **Self-Signed Certificates**: Auto-generate self-signed certs with `selfsigned` option
- **Custom Certificates**: Load custom cert/key files for production deployments
- **CORS Enabled**: Allow all origins for localhost-only development
### REST API
- **Health Check**: `GET /api/health` returns server status
- **Config Routes**: `GET /api/config` and `POST /api/config` for configuration management
- **Process Routes**: `GET /api/processes` returns all active sessions with process details
- **Cache Streaming**: Request cache stream for specific session via WebSocket `stream-cache` event
### UI Features (Implemented)
- **React Router**: Client-side routing with `/` (ProcessMonitor), `/config` (ConfigEditor), `/logs` (LogViewer) ✅
- **Global Modals**: Permission approval modal integrated in App.tsx with WebSocket subscriptions ✅
- **Layout Component**: Shared navigation with 3 pages (Processes, Configuration, Logs) ✅
- **TanStack Query**: Optimized data fetching with caching and refetch strategies
- **Responsive Design**: Mobile-friendly UI (framework in place, styling in progress)
- **WebSocket Hooks**: Callback refs pattern in `useWebSocket` for permission and log subscriptions ✅
---
## Programmatic API (Phase 3) 🔮
**Status**: Express foundation in place, RESTful endpoints planned
### RESTful Endpoints (Planned)
- **Team Operations**: `POST /api/teams/{team}/tell`, `POST /api/teams/{team}/wake`, `DELETE /api/teams/{team}`
- **Session Management**: `GET /api/sessions`, `GET /api/sessions/{sessionId}`, `DELETE /api/sessions/{sessionId}`
- **Process Control**: `GET /api/processes`, `POST /api/processes/{team}/reboot`, `POST /api/processes/{team}/compact`
- **Configuration**: `GET /api/config`, `PUT /api/config`, `POST /api/config/teams`
- **Notifications**: `GET /api/notifications/{team}`, `POST /api/notifications`, `DELETE /api/notifications/{id}`
### WebSocket Streaming (Planned)
- **Live Message Stream**: Subscribe to team message streams with `socket.on('messages', ...)`
- **Process Events**: Real-time process-spawned, process-terminated, process-error events
- **Cache Streaming**: Stream full cache history for session with backpressure handling
- **Presence**: Track connected clients per team with join/leave events
### Authentication (Planned)
- **API Keys**: Bearer token authentication with configurable scopes (read, write, admin)
- **Key Management**: `POST /api/auth/keys` to create API keys with expiration
- **Rate Limiting**: Per-key rate limits to prevent abuse (100 req/min default)
- **Audit Logging**: Track all API calls with key, endpoint, timestamp, result
### Job Queue (Planned)
- **Async Operations**: Queue long-running operations (wake_all, multi-team messaging) with job IDs
- **Job Status**: `GET /api/jobs/{jobId}` returns pending, running, completed, failed
- **Job Cancellation**: `DELETE /api/jobs/{jobId}` to cancel running jobs
- **Job History**: Persist job results for 7 days with automatic cleanup
### Official SDKs (Planned)
- **TypeScript SDK**: Fully-typed client with `IrisClient` class, auto-retry, and error handling
- **Python SDK**: `iris-mcp-client` package with async/await support and type hints
- **REST Client Examples**: cURL examples for all endpoints in API documentation
- **WebSocket Examples**: Socket.IO client examples in JS and Python
---
## CLI Interface (Phase 4) 🔮
**Status**: Foundation in place (Ink + Commander installed), command structure planned
### Current CLI (Setup Commands) ✅
- **add-team**: Add a new team to configuration (`iris-mcp add-team <name> <path>`)
- **install**: Install Iris MCP server to Claude Code's MCP config (`iris-mcp install`)
- **uninstall**: Remove Iris MCP from Claude Code's MCP config (`iris-mcp uninstall`)
### Planned Interactive CLI
**Message Commands**:
- `iris ask <team> <question>`: Ask a team a question and wait for response
- `iris tell <team> <message>`: Send a message to a team
- `iris notify <team> <message>`: Send fire-and-forget notification
**Team Commands**:
- `iris teams`: List all configured teams with status
- `iris status [team]`: Check team process status and metrics
- `iris wake <team>`: Wake up a team process
- `iris sleep <team>`: Put a team to sleep
- `iris fork <team>`: Launch interactive terminal session
**Session Commands**:
- `iris sessions`: List all active sessions with message counts
- `iris compact <team>`: Compress session history
- `iris clear <team>`: Create fresh session (reboot)
- `iris logs <team>`: View recent logs for team process
**Interactive Shell**:
- `iris shell`: Launch interactive REPL with autocomplete
- `> wake frontend`: Execute commands without `iris` prefix
- `> ask backend "What's the API version?"`: Multi-word arguments
- Tab completion for team names and commands
### Rich Terminal UI (Ink Components)
- **Live Dashboard**: `iris monitor` shows live process status in terminal UI
- **Progress Indicators**: Spinners for async operations with estimated time
- **Color-Coded Output**: Team colors from config for visual distinction
- **Table Formatting**: Pretty-printed tables for team/session lists
- **Syntax Highlighting**: JSON output with colors for readability
---
## Intelligence Layer (Phase 5) 🔮
**Status**: Event system foundation in place, autonomous features planned
### Loop Detection
- **Message Pattern Analysis**: Detect cyclical message patterns (A→B→A→B...)
- **Infinite Loop Prevention**: Automatically break loops after N iterations
- **Loop Alerts**: Notify when loop detected with suggested resolution
- **Deadlock Detection**: Identify teams waiting on each other with timeout warnings
### Destructive Action Prevention
- **Risk Assessment**: Analyze tool calls for potential destructive operations
- **Confirmation Prompts**: Require human approval for high-risk actions (delete, mass-reboot)
- **Rollback Capability**: Track state changes to enable rollback on error
- **Dry-Run Mode**: Simulate destructive actions without execution
### Pattern Recognition
- **Common Workflows**: Learn frequently-used message sequences (frontend→backend→database)
- **Workflow Shortcuts**: `iris workflow deploy` executes learned multi-step pattern
- **Anomaly Detection**: Alert when teams deviate from normal communication patterns
- **Performance Optimization**: Suggest workflow improvements based on observed bottlenecks
### Self-Aware Coordination
- **Meta-Cognition**: Iris analyzes its own coordination patterns and suggests improvements
- **Auto-Scaling**: Dynamically adjust `maxProcesses` based on demand
- **Proactive Waking**: Wake teams before they're needed based on predicted usage
- **Context Sharing**: Automatically share relevant context between teams without explicit tells
---
## Performance Metrics
### Process Pool
- **Cold Start**: 7-14 seconds (session creation + process spawn)
- **Warm Start**: 0.5-2 seconds (process reuse, 10-20x faster)
- **Session Lookup**: ~1ms (60-second cache hit)
- **LRU Eviction**: <10ms (terminate + cleanup)
### Memory Usage
- **10 Processes**: 600 MB - 1.25 GB RAM
- **Session Database**: ~2 MB per 10,000 sessions
- **Notification Queue**: ~5 MB per 10,000 notifications
- **Message Cache**: ~1-5 MB per session (depends on message count)
### Test Suite
- **Unit Tests**: 203 tests passing in <2 seconds (mocked)
- **Integration Tests**: 85% faster (7min → 1min) with `beforeAll` optimization
- **Coverage**: V8 coverage provider with branch/line/function metrics
### Remote Execution
- **SSH Connection**: ~100-500ms (network latency dependent)
- **Reverse Tunnel**: ~100ms one-time overhead per connection
- **MCP Tool Latency**: ~50-200ms (network latency + SSH encryption)
- **Permission Approval**: ~10-50ms (cached after first approval)
---
## Feature Comparison
### Iris MCP vs Other Orchestration Frameworks
| Feature | Iris MCP | Symphony of One | Claude-Flow | Agent-MCP | Others |
|---------|-----------|-----------------|-------------|-----------|--------|
| **Cross-Project Communication** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Independent Team Contexts** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Direct Agent-to-Agent Messaging** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Per-Team MCP Server Access** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Zero Shared State** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Natural Language Coordination** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Remote Execution via SSH** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Reverse MCP Tunnel** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Persistent Session Context** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Process Pool Management** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **OpenTelemetry Integration** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Distributed Tracing** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Prometheus Metrics** | ✅ | ❌ | ❌ | ❌ | ❌ |
| **Web Dashboard** | 🚧 | ❌ | ❌ | ❌ | ❌ |
| **RESTful API** | 🔮 | ❌ | ❌ | ❌ | ❌ |
| **Interactive CLI** | 🔮 | ❌ | ❌ | ❌ | ❌ |
| **Intelligence Layer** | 🔮 | ❌ | ❌ | ❌ | ❌ |
**Legend**: ✅ Implemented | 🚧 In Progress | 🔮 Planned | ❌ Not Available
### Unique Capabilities
1. **True Cross-Codebase Coordination**: Only Iris enables Claude instances across completely independent projects to communicate
2. **Context Isolation**: Each team maintains separate `.claude/` config, MCP servers, dependencies, and git repos
3. **Bidirectional Remote Orchestration**: Remote teams can call back to local Iris via SSH reverse tunneling
4. **Session Persistence**: Full conversation history maintained across process restarts
5. **Warm Process Pooling**: 10-20x performance improvement via process reuse
6. **Production-Grade Observability**: OpenTelemetry integration with distributed tracing, metrics, and log correlation
7. **Multi-Interface Access**: MCP tools, web dashboard, REST API, CLI (current + planned)
---
## Technology Stack
### Core (Phase 1)
- **Runtime**: Node.js 18+ with ES2022 modules
- **Language**: TypeScript 5.7+ with strict mode
- **MCP Protocol**: `@modelcontextprotocol/sdk` for MCP server implementation
- **Process Management**: Node.js `child_process` with stdio communication
- **Database**: `better-sqlite3` for session store and notification queue
- **Validation**: Zod schemas for configuration and input validation
- **Configuration**: YAML parsing with `yaml` package, environment variable interpolation
- **Observability**: `@jenova-marie/wonder-logger` with Pino + OpenTelemetry
- **Tracing**: OpenTelemetry SDK with OTLP, Jaeger exporters
- **Metrics**: Prometheus pull endpoint + OTLP push
- **Testing**: Vitest with V8 coverage provider
### Dashboard (Phase 2)
- **Frontend**: React 18.2+ with TypeScript
- **Routing**: React Router 7.1+ for SPA navigation
- **State**: TanStack Query (React Query) for server state
- **Real-Time**: Socket.IO for WebSocket connections
- **Backend**: Express 5.0+ with CORS middleware
- **HTTPS**: Self-signed certificates via `selfsigned` package
- **Build**: Vite for client bundling with HMR
### API (Phase 3)
- **HTTP Server**: Express (already installed for Dashboard)
- **WebSocket**: Socket.IO (already installed for Dashboard)
- **Authentication**: JWT tokens with `jsonwebtoken` (planned)
- **Rate Limiting**: `express-rate-limit` (planned)
- **API Docs**: OpenAPI 3.0 spec (planned)
### CLI (Phase 4)
- **Framework**: Commander.js for argument parsing
- **Terminal UI**: Ink 5.0+ (React for terminals)
- **Styling**: Chalk for colors (via Ink)
- **Prompts**: Inquirer.js for interactive questions (planned)
### Remote Execution
- **SSH Client**: OpenSSH CLI via `child_process` (default, implemented)
- **Transport Abstraction**: Interface-based design for LocalTransport vs SSHTransport
---
## Documentation
- **[GETTING_STARTED.md](../GETTING_STARTED.md)**: Installation and quick start guide
- **[ARCHITECTURE.md](./ARCHITECTURE.md)**: System design and component interaction
- **[CONFIG.md](./CONFIG.md)**: Complete configuration reference with YAML format and env vars
- **[ACTIONS.md](./ACTIONS.md)**: Complete MCP tools API reference
- **[REMOTE.md](./REMOTE.md)**: Remote execution via SSH documentation
- **[REVERSE_MCP.md](./REVERSE_MCP.md)**: Bidirectional tunnel architecture and security
- **[SESSION.md](./SESSION.md)**: Session management deep dive
- **[PROCESS_POOL.md](./PROCESS_POOL.md)**: Process pool management and LRU eviction
- **[CACHE.md](./CACHE.md)**: Message cache system with RxJS observables
- **[BREAKING.md](./BREAKING.md)**: Migration guide for breaking changes
---
## Related Resources
- **GitHub Repository**: https://github.com/jenova-marie/iris-mcp
- **NPM Package**: https://www.npmjs.com/package/@jenova-marie/iris-mcp
- **Issue Tracker**: https://github.com/jenova-marie/iris-mcp/issues
- **Contributing**: [CONTRIBUTING.md](../CONTRIBUTING.md)
- **License**: MIT
---
## Tech Writer Notes
**Coverage Areas:**
- All implemented and planned Iris MCP features across 5 phases
- Configuration management (YAML format, environment variables, grantPermission field)
- MCP tools inventory and categorization
- Performance metrics and benchmarks
- Technology stack breakdown by phase
- Feature comparison with other frameworks
- Documentation cross-references
**Keywords:** features, capabilities, MCP tools, configuration, YAML, environment variables, grantPermission, process pool, remote execution, observability, dashboard, API, CLI, intelligence layer, performance metrics, technology stack
**Last Updated:** 2025-10-18
**Change Context:** Updated MCP Tools section with new tool names (v3.0). Renamed: team_tell → send_message, team_quick_tell → quick_message, team_isAwake → team_status, team_report → session_report, team_teams → list_teams, team_debug → get_logs, team_cancel → session_cancel, team_delete → session_delete, team_fork → session_fork, team_clear → session_reboot. Added semantic aliases: ask_message (for send_message), team_launch (for team_wake). Added get_date tool. Removed team_compact (incomplete implementation). Updated tool count from 15 to 17 tools. Previous update (2025-10-17): Updated permission approval system status to ✅ fully implemented with dashboard UI.
**Related Files:** ACTIONS.md (complete tool API reference), NOMENCLATURE.md (core concepts), PERMISSIONS.md (permission system details), DASHBOARD.md (UI documentation), CONFIG.md (configuration), SESSION.md (debug info fields), GETTING_STARTED.md (quick start), README.md (overview), ARCHITECTURE.md (system design)
---
**Document Version**: 2.1
**Last Updated**: October 2025
**Status Key**: ✅ Implemented | 🚧 In Progress | 🔮 Planned