# CrewAI + Pomera MCP Integration
Use Pomera's 22+ text processing tools inside [CrewAI](https://crewai.com) agent workflows.
## Architecture
```
┌─────────────────────────────────────────────────────────────────┐
│ CrewAI Python Script │
│ MCPServerStdio(command="npx", args=["pomera-mcp"]) │
│ │ │
│ └── spawns Pomera MCP (same as IDE does) │
│ ↑↓ stdio │
│ Agent uses pomera_web_search, pomera_notes, etc. │
└─────────────────────────────────────────────────────────────────┘
```
**Same pattern as IDE** - CrewAI spawns MCP server via stdio.
---
## Setup
```bash
pip install crewai crewai-tools
```
---
## Example Crews
### Research Crew
```python
# crews/research_crew.py
from crewai import Agent, Task, Crew, Process
from crewai.mcp import MCPServerStdio
import sys
# Pomera MCP server (spawned like IDE does)
pomera_mcp = MCPServerStdio(
command="npx",
args=["pomera-mcp"],
env={"OPENAI_API_KEY": "your-key"} # if needed
)
researcher = Agent(
role="Research Analyst",
goal="Find comprehensive information on {topic}",
backstory="Expert researcher using multiple search engines",
mcps=[pomera_mcp],
verbose=True
)
analyzer = Agent(
role="Content Analyzer",
goal="Synthesize research into actionable insights",
backstory="Analyst who transforms data into insights",
mcps=[pomera_mcp],
verbose=True
)
search_task = Task(
description="Search for {topic} using pomera_web_search with tavily",
expected_output="List of relevant sources with summaries",
agent=researcher
)
analyze_task = Task(
description="Analyze findings using pomera_ai_tools deepreasoning",
expected_output="Structured analysis with key insights",
agent=analyzer,
context=[search_task]
)
crew = Crew(
agents=[researcher, analyzer],
tasks=[search_task, analyze_task],
process=Process.sequential,
verbose=True
)
if __name__ == "__main__":
topic = sys.argv[1] if len(sys.argv) > 1 else "AI orchestration"
result = crew.kickoff(inputs={"topic": topic})
print(result)
```
### Blog Crew
```python
# crews/blog_crew.py
from crewai import Agent, Task, Crew, Process
from crewai.mcp import MCPServerStdio
import sys
pomera_mcp = MCPServerStdio(command="npx", args=["pomera-mcp"])
researcher = Agent(
role="SEO Researcher",
goal="Find trending keywords for {niche}",
backstory="SEO expert",
mcps=[pomera_mcp]
)
writer = Agent(
role="Blog Writer",
goal="Write engaging blog posts",
backstory="Professional blogger",
mcps=[pomera_mcp]
)
keyword_task = Task(
description="Find top keywords for {niche} using pomera_web_search",
expected_output="10 keywords with search intent",
agent=researcher
)
write_task = Task(
description="Write 1500-word blog post targeting the keywords",
expected_output="SEO-optimized blog post in markdown",
agent=writer,
context=[keyword_task],
output_file="blog_output.md"
)
crew = Crew(agents=[researcher, writer], tasks=[keyword_task, write_task])
if __name__ == "__main__":
niche = sys.argv[1] if len(sys.argv) > 1 else "cycling"
crew.kickoff(inputs={"niche": niche})
print("Blog saved to blog_output.md")
```
---
## Run From IDE
```bash
python crews/research_crew.py "AI agent frameworks"
python crews/blog_crew.py "road cycling"
```
---
## Pomera Tools Available
| Tool | Use |
|------|-----|
| `pomera_web_search` | 5 search engines |
| `pomera_ai_tools` | 11 AI providers |
| `pomera_notes` | Persistent memory |
| `pomera_read_url` | Web scraping |
See [Tools Documentation](tools/INDEX.md) for the full list.
---
## Save Results to Notes
```python
save_task = Task(
description="Save analysis to pomera_notes with title 'Research/{topic}'",
expected_output="Note ID confirmation",
agent=analyzer,
context=[analyze_task]
)
```
