AgentPipe orchestrates multi-agent conversations by enabling different AI CLI tools to communicate in shared rooms. It benefits operations teams by automating complex workflows. The app connects to tools like Claude Code, Gemini, and Qwen, streamlining AI-driven tasks.
git clone https://github.com/kevinelliott/agentpipe.gitAgentPipe is a CLI and TUI application that enables multiple AI agents to communicate in shared conversation rooms using a standardized interaction pattern. It supports 14+ AI agents including Claude, Cursor, Gemini, Qwen, Ollama, and OpenRouter, allowing teams to automate complex workflows through orchestrated multi-agent conversations. The tool offers three conversation modes—round-robin, reactive, and free-form—with real-time metrics, token tracking, and cost analysis. Built-in features include Prometheus metrics for observability, conversation export (JSON, Markdown, HTML), rate limiting, retry logic, and Docker support. Operations and development teams benefit from automated task execution, interactive participation in agent conversations, and comprehensive logging through the enhanced TUI interface with agent status indicators and cost analytics.
[{"step":"Install AgentPipe and required CLI tools (e.g., `pip install agentpipe`, ensure Claude Code, Qwen, or other tools are installed).","action":"Run `agentpipe init` to set up a new room and configure the agents."},{"step":"Define the workflow in the AgentPipe room configuration file (e.g., `agentpipe_config.yaml`). Specify the agents, their roles, and the sequence of tasks.","action":"Use the prompt template to generate the configuration. Example: `agentpipe_config.yaml` includes `agents: [claude_code, qwen]` and `workflow: [refactor_code, validate_quality, generate_tests]`."},{"step":"Start the pipeline with `agentpipe run --config agentpipe_config.yaml`.","action":"Monitor the room’s output in the terminal or AgentPipe’s dashboard for real-time progress."},{"step":"Review the final output and validation report.","action":"Address any errors or bottlenecks flagged by the agents. For example, if Agent B flags missing tests, use Agent A to generate them automatically."},{"step":"Integrate the results into your workflow (e.g., commit the code, trigger CI/CD, or notify stakeholders).","action":"Use the `agentpipe export` command to save the output to a file or repository."}]
Automate complex software development workflows with multiple coding agents collaborating
Run parallel AI agents for code review, debugging, and optimization tasks
Track token usage and costs across multiple AI providers in a single conversation
Orchestrate multi-agent problem-solving with round-robin or reactive conversation modes
No install command available. Check the GitHub repository for manual installation instructions.
git clone https://github.com/kevinelliott/agentpipeCopy the install command above and run it in your terminal.
Launch Claude Code, Cursor, or your preferred AI coding agent.
Use the prompt template or examples below to test the skill.
Adapt the skill to your specific use case and workflow.
Set up an AgentPipe room with [TOOL_A] and [TOOL_B] to automate [TASK]. Configure the agents to follow this workflow: [STEP_1], [STEP_2], and [STEP_3]. Ensure [TOOL_A] handles [SPECIFIC_ACTION] and [TOOL_B] validates [OUTPUT_CRITERIA]. Run the pipeline and provide a summary of the results, including any errors or bottlenecks encountered.
AgentPipe Room: **Code Review & Documentation Pipeline** **Agents Assigned:** - **Claude Code (Agent A):** Primary executor for code generation and refactoring. - **Qwen (Agent B):** Validator for code quality, documentation accuracy, and compliance with [COMPANY_STYLE_GUIDE]. **Workflow Execution:** 1. **Input:** User provided a Python script (`data_processor.py`) with a bug in the `calculate_average()` function and missing docstrings. 2. **Agent A (Claude Code):** - Detected the bug in `calculate_average()` (incorrect handling of empty lists). - Refactored the function to include a `ZeroDivisionError` check and added type hints. - Generated a draft docstring for the function. - Output: Updated `data_processor.py` with fixes. 3. **Agent B (Qwen):** - Validated the refactored code against [COMPANY_STYLE_GUIDE] (PEP 8 compliance, docstring format). - Flagged missing unit tests for the new function. - Suggested adding a `pytest` test case for `calculate_average()`. - Output: Approved the code with annotations for missing tests. 4. **Agent A (Claude Code):** - Generated a `test_data_processor.py` file with a test case for `calculate_average()`. - Verified the test passed locally. 5. **Final Output:** - **Code:** `data_processor.py` (refactored, bug-free, documented). - **Tests:** `test_data_processor.py` (new test case added). - **Validation Report:** Qwen confirmed 100% PEP 8 compliance and docstring accuracy. **Errors/Bottlenecks:** None. The pipeline executed in 4 minutes with no manual intervention required. Agent B’s suggestion for unit tests was incorporated seamlessly by Agent A. **Next Steps:** - Merge the changes into the main branch. - Run the full test suite in CI/CD to ensure no regressions.
Google's multimodal AI model and assistant
AI assistant built for thoughtful, nuanced conversation
Get more done every day with Microsoft Teams – powered by AI
Automate your spreadsheet tasks with AI power
Agentic AI Workflow platform
Connected workspace for docs, wikis, and projects
Take a free 3-minute scan and get personalized AI skill recommendations.
Take free scan