AgentPipe 🚀

AgentPipe is a powerful CLI and TUI application that orchestrates conversations between multiple AI agents. It allows different AI CLI tools (like Claude, Cursor, Gemini, Qwen, Ollama) to communicate with each other in a shared "room", creating dynamic multi-agent conversations with real-time metrics, cost tracking, and interactive user participation.

Screenshots

Enhanced TUI Interface

Enhanced TUI with multi-panel layout: agent list with status indicators, conversation view with metrics, statistics panel showing turns and total cost, configuration panel, and user input area

Console/CLI Interface

CLI output showing color-coded agent messages with agent type indicators (e.g., "Alice (qoder)"), HOST vs SYSTEM distinction, and inline metrics display

Supported AI Agents

All agents now use a standardized interaction pattern with structured three-part prompts, message filtering, and comprehensive logging for reliable multi-agent conversations.

• ✅ Amp (Sourcegraph) - Advanced coding agent with autonomous reasoning ⚡ Thread-optimized
• ✅ Claude (Anthropic) - Advanced reasoning and coding
• ✅ Codex (OpenAI) - Code generation specialist (non-interactive exec mode)
• ✅ Copilot (GitHub) - Terminal-based coding agent with multiple model support
• ✅ Continue - AI coding assistant with TUI and headless modes for development workflows
• ✅ Crush (Charm/Charmbracelet) - Terminal-first AI coding assistant with multi-provider support
• ✅ Cursor (Cursor AI) - IDE-integrated AI assistance
• ✅ Factory (Factory.ai) - Agent-native software development with Droid (non-interactive exec mode)
• ✅ Gemini (Google) - Multimodal understanding
• ✅ Groq - Fast AI code assistant powered by Groq LPUs (Lightning Processing Units)
• ✅ Kimi (Moonshot AI) - Interactive AI agent with advanced reasoning (interactive-first CLI)
• ✅ OpenCode (SST) - AI coding agent built for the terminal (non-interactive run mode)
• ✅ OpenRouter - Unified API access to 400+ models from multiple providers (API-based, no CLI required) 🌐 API-based
• ✅ Qoder - Agentic coding platform with enhanced context engineering
• ✅ Qwen (Alibaba) - Multilingual capabilities
• ✅ Ollama - Local LLM support (planned)

Features

Core Capabilities

• Multi-Agent Conversations: Connect multiple AI agents in a single conversation
• Multiple Conversation Modes:
  • round-robin: Agents take turns in a fixed order
  • reactive: Agents respond based on conversation dynamics
  • free-form: Agents participate freely as they see fit
• Flexible Configuration: Use command-line flags or YAML configuration files

Enhanced TUI Interface

• Multi-panel layout with dedicated sections for agents, chat, stats, and config
• Color-coded agent messages with unique colors per agent
• Agent type indicators showing agent type in parentheses (e.g., "Alice (qoder)")
• Branded sunset logo with gradient colors
• Real-time agent activity indicators (🟢 active/responding, ⚫ idle)
• Inline metrics display (response time in seconds, token count, cost)
• Conversation search (Ctrl+F) with n/N navigation through results
• Agent filtering via slash commands (/filter, /clear)
• Help modal (?) showing all keyboard shortcuts
• Topic panel showing initial conversation prompt
• Statistics panel with turn counters and total conversation cost
• Configuration panel displaying all active settings and config file path
• Interactive user input panel for joining conversations
• Smart message consolidation (headers only on speaker change)
• Proper multi-paragraph message formatting

Production Features

• Prometheus Metrics: Comprehensive observability with 10+ metrics types
  • Request rates, durations, errors
  • Token usage and cost tracking
  • Active conversations, retry attempts, rate limit hits
  • HTTP server with /metrics, /health, and web UI endpoints
  • Ready for Grafana dashboards and alerting
• Conversation Management:
  • Save/resume conversations from state files
  • Export to JSON, Markdown, or HTML formats
  • Automatic chat logging to ~/.agentpipe/chats/
• Reliability & Performance:
  • Rate limiting per agent with token bucket algorithm
  • Retry logic with exponential backoff (configurable)
  • Structured error handling with error types
  • Config hot-reload for development workflows
• Middleware Pipeline: Extensible message processing
  • Built-in: logging, metrics, validation, sanitization, filtering
  • Custom middleware support for transforms and filters
  • Error recovery and panic handling
• Docker Support: Multi-stage builds, docker-compose, production-ready
• Health Checks: Automatic agent health verification before conversations
• Agent Detection: Built-in doctor command to check installed AI CLIs
• Customizable Agents: Configure prompts, models, and behaviors for each agent

What's New

See CHANGELOG.md for detailed version history and release notes.

Latest Release: v0.6.0 - OpenRouter API Support

What's New in v0.6.0:

🌐 OpenRouter API Support - First API-Based Agent:

• New Agent Type: Direct API integration without CLI dependencies
  • Access 400+ models from multiple providers through a unified API
  • No CLI installation required - just set OPENROUTER_API_KEY
  • Support for models from Anthropic, OpenAI, Google, DeepSeek, and more
  • Real-time token usage and accurate cost tracking from API responses
  • Streaming and non-streaming message support
  • Smart model matching with provider registry integration
• Example Configurations:
  • examples/openrouter-conversation.yaml - Multi-provider conversation
  • examples/openrouter-solo.yaml - Single agent testing
• Foundation for Future API Agents: Paves the way for direct Anthropic API, Google AI API, etc.

Previous Release - v0.4.9 (2025-10-25): Crush CLI support

• Full support for Charm's Crush CLI with multi-provider capabilities

Previous Release - v0.4.8 (2025-10-25): Fixed GitHub API rate limiting

• PyPI integration for Kimi, npm registry for Qwen
• No more 403 rate limit errors

Previous Release - v0.4.7 (2025-10-25): Improved Kimi version detection Previous Release - v0.4.6 (2025-10-25): Groq Code CLI support

• Dedicated security workflows (Trivy and CodeQL)
• Enhanced README badges with downloads and stars metrics
• Fixed Windows test failures for platform-specific installations

Previous Release - v0.4.3 (2025-10-25): Kimi CLI agent support

• Full support for Kimi CLI from Moonshot AI
• Installation via uv tool install kimi-cli (requires Python 3.13+)

Previous Release - v0.4.2 (2025-10-24): Qoder installation improvements

• Added --force flag to Qoder install and upgrade commands

Previous Release - v0.4.1 (2025-10-22): Conversation summarization & unique agent IDs

• AI-generated summaries at conversation completion
• Unique agent IDs for multiple agents of same type
• Local event storage to ~/.agentpipe/events/

Previous Release - v0.4.0: Bridge connection events and cancellation detection Previous Release - v0.3.0: Real-time conversation streaming to AgentPipe Web Previous Release - v0.2.2: JSON output support for agents list command Previous Release - v0.2.1: OpenCode agent and improved package management Previous Release - v0.2.0: Agent upgrade command and automated version detection

Installation

Using Homebrew (macOS/Linux)

brew tap kevinelliott/tap
brew install agentpipe

Using the install script

curl -sSL https://raw.githubusercontent.com/kevinelliott/agentpipe/main/install.sh | bash

Using Go

go install github.com/kevinelliott/agentpipe@latest

Build from source

git clone https://github.com/kevinelliott/agentpipe.git
cd agentpipe

# Build only
go build -o agentpipe .

# Or build and install to /usr/local/bin (requires sudo on macOS/Linux)
sudo make install

# Or install to custom location (e.g., ~/.local/bin, no sudo needed)
make install PREFIX=$HOME/.local

Prerequisites

AgentPipe requires at least one AI CLI tool to be installed:

• Amp CLI - amp ⚡ Optimized
  • Install: npm install -g @sourcegraph/amp or see installation guide
  • Authenticate: Follow Amp documentation
  • Features: Autonomous coding, IDE integration, complex task execution
  • Thread Management: AgentPipe uses Amp's native threading to maintain server-side conversation state
  • Smart Filtering: Only sends new messages from other agents, reducing API costs by 50-90%
  • Structured Context: Initial prompts are delivered in a clear, three-part structure
• [Claude CLI](https