CrewAI MCP Orchestrator
This MCP server enables an LLM to act as a CrewAI orchestrator, managing the full lifecycle of agent crews — from creation and templating to execution, debugging, and optimization — via 15 tools and a RAG knowledge engine.
Project Management
Create new CrewAI crew or flow projects, scaffold directory structure, pyproject.toml, and YAML configs (
crewai_create_project)Install dependencies and optional extra packages (
crewai_install_deps)Inspect a project's dependencies, YAML configs, and source files (
crewai_project_info)
Templating
Apply prebuilt crew templates (e.g., 5-agent CyberOps MVP) to instantly scaffold a full team (
crewai_apply_template)
Agents & Tasks
Define agents with custom roles, goals, backstories, LLMs, and tools (
crewai_define_agent)Define tasks with descriptions, expected outputs, assigned agents, and context dependencies (
crewai_define_task)Modify agent parameters in crew.py (LLM, tools, custom constructor args) (
crewai_edit_crew_py)
Execution & Flows
Kick off a crew with optional input variables (
crewai_kickoff)Generate an interactive HTML visualization of a flow's states and transitions (
crewai_flow_plot)Execute a flow project with optional inputs (
crewai_flow_run)
Knowledge & Memory
Query a RAG-indexed database of 266+ CrewAI documentation articles (
crewai_query_knowledge)Reset or check the status of project memory (
crewai_manage_memory)
Debugging & Optimization
Test crew output quality over multiple iterations using an LLM judge (
crewai_test_crew)Run human-in-the-loop training to improve agent prompts (
crewai_train_crew)Replay a specific failed task by ID without restarting the entire crew (
crewai_replay_task)
It connects via stdio or SSE to any MCP client (Claude Desktop, Cursor, Roo Code, etc.).
Allows orchestration of CrewAI multi-agent projects, including creating projects, defining agents and tasks, running flows, and querying documentation via RAG.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@CrewAI MCP OrchestratorCreate a new crew project for content writing"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
🚀 CrewAI MCP Orchestrator
MCP server that turns any LLM into a CrewAI orchestrator. 15 tools, prebuilt crew templates, and RAG engine with 266+ indexed docs.
📖 Documentation: English · Español
⚡ Install
git clone https://github.com/ssolis-ti/crewai-mcp-hq.git
cd crewai-mcp-hq
uv syncRelated MCP server: Code-MCP
🔌 Connect to MCP Clients
Hermes Agent
hermes mcp add crewai-orchestrator \
--command "/path/to/crewai-mcp-hq/.venv/Scripts/python.exe"
--args "-X utf8 -m crewai_mcp.server"Claude Desktop / Cursor / Roo Code
{
"mcpServers": {
"crewai-orchestrator": {
"command": "/path/to/crewai-mcp-hq/.venv/bin/python",
"args": ["-m", "crewai_mcp.server"],
"cwd": "/path/to/crewai-mcp-hq",
"env": { "CREWAI_MCP_TRANSPORT": "stdio" }
}
}
}Docker (SSE)
docker-compose up -d
# Available at http://localhost:8808/sse🧰 Tools (15)
Domain | Tools |
Projects |
|
Templates |
|
Agents & Tasks |
|
Flows |
|
Knowledge |
|
Observability |
|
🧩 Prebuilt Crew Templates
Deploy a full team in one call — no per-agent setup:
crewai_create_project(name="my-mvp", project_type="crew")
crewai_apply_template(project_name="my-mvp", template_name="cyberops")
# agents.yaml, tasks.yaml, and crew.py ready to runCyberOps — MVP Development Team
5-agent sequential crew. Input: project description. Output: PRD + architecture + code + docs + QA.
Agent | Role | Configurable |
PRD_Architect | Requirements & user stories | LLM, tools, max_iter |
System_Designer | Architecture (ADRs, C4, API) | LLM, tools, max_iter |
AI_Developer | AI-first code (<100 lines/file) | LLM, tools, max_iter |
Doc_Engineer | LLM-optimized documentation | LLM, tools, max_iter |
QA_Reviewer | Quality audit & traceability | LLM, tools, max_iter |
🗺️ Deployment Workflow (with your AI assistant)
The logical order to deploy a team of agents using the MCP. Just tell your assistant "I need a team for X" and it handles the rest:
1. CREATE crewai_create_project("my-team", "crew")
↓
2. TEMPLATE crewai_apply_template("my-team", "cyberops")
↓
3. INSTALL crewai_install_deps("my-team")
↓
4. KICKOFF crewai_kickoff("my-team", inputs={...})
↓
5. ITERATE crewai_test_crew / crewai_replay_task / crewai_train_crewStep-by-step with your AI assistant
Step | What you say | Tool called |
Research | "I need a team to build [project]" |
|
Scaffold | "Create the project" |
|
Template | "Apply CyberOps template" |
|
Customize | "Change AI_Developer to use gpt-4" |
|
Install | "Install dependencies" |
|
Run | "Execute the crew" |
|
Debug | "QA agent failed — retry it" |
|
Improve | "Test and train" |
|
Building a custom team from scratch
No prebuilt template? Define agents and tasks one by one:
1. CREATE crewai_create_project("my-custom", "crew")
2. AGENTS crewai_define_agent("my-custom", "researcher", role="...")
crewai_define_agent("my-custom", "writer", role="...")
3. TASKS crewai_define_task("my-custom", "research", agent="researcher")
crewai_define_task("my-custom", "write", agent="writer")
4. INSTALL crewai_install_deps("my-custom")
5. KICKOFF crewai_kickoff("my-custom", inputs={...})📚 Documentation Resources
URI | Content |
| 266+ docs across 31 categories |
| Specific documentation pages |
| Keyword search |
| Agent, crew & flow templates |
| Full crew templates (CyberOps + extensible) |
🛡️ Robustness
Auto-patch versions:
crewai createoutputs pre-release pins → auto-patched to>=1.14.0Name normalization: hyphens/underscores handled transparently
Timeouts on all subprocess calls: 120s–1200s depending on operation
Standardized CLI: always
uv run crewai, no PATH dependency
📁 Structure
src/crewai_mcp/
├── server.py ← Entry point (stdio/sse/streamable-http)
├── resources/ ← Docs, templates, prebuilt crews
├── tools/ ← 15 tools + shared utils.py
├── prompts/ ← Guided workflows (design_crew, debug_crew)
└── knowledge/ ← ChromaDB indexer + retriever📖 Documentación en Español
La documentación de CrewAI está disponible en inglés en docs.crewai.com. Para usar el MCP en español:
El motor RAG indexa docs en inglés pero responde preguntas en cualquier idioma
Los templates de crews aceptan descripciones de proyecto en español
Las herramientas retornan mensajes en inglés; el LLM que consume el MCP traduce al contexto del usuario
Guías rápidas en español:
Guía | Descripción |
Clonar, instalar dependencias, conectar a tu IDE | |
Equipo de 5 agentes para crear MVPs desde cero | |
Referencia completa de las 15 herramientas | |
|
📝 License
MIT
Maintenance
Latest Blog Posts
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/ssolis-ti/crewai-mcp-hq'
If you have feedback or need assistance with the MCP directory API, please join our Discord server