Frequently Asked Questions
Common questions about RunAgent
General Questions
RunAgent is a platform for deploying AI agents to production. It provides:
- Framework-agnostic deployment
- Automatic scaling and monitoring
- Simple CLI and SDK interfaces
- Support for all major AI frameworks
Think of it as “Vercel for AI agents” - deploy with one command and let us handle the infrastructure.
Currently supported:
- LangGraph
- CrewAI
- Agno
- Letta
- Custom Python frameworks
Coming soon:
- AutoGen
- BabyAGI
- AgentGPT
You can use any Python-based framework with our custom framework option.
Local deployment: Free forever
Cloud deployment (coming soon):
- Free tier: 1,000 requests/month
- Pro: $29/month for 50,000 requests
- Enterprise: Custom pricing
See pricing page for details.
Getting Started
Simply use pip:
pip install runagent
For development with all dependencies:
pip install "runagent[dev]"
No! RunAgent works with your existing code. Just:
- Add a
runagent.config.jsonfile - Define your entrypoints
- Deploy
Your agent logic remains unchanged.
Entrypoints are functions RunAgent calls to interact with your agent:
# Standard entrypoint
def invoke(input_data: dict) -> dict:
return {"result": "processed"}
# Streaming entrypoint
def stream(input_data: dict):
for chunk in process():
yield chunk
Technical Questions
RunAgent supports Python 3.8 and above:
- Python 3.8
- Python 3.9
- Python 3.10
- Python 3.11
- Python 3.12
Yes! Add any dependencies to your requirements.txt:
openai>=1.0.0
requests>=2.31.0
pandas>=2.0.0
your-custom-package>=1.0.0
RunAgent will install them during deployment.
Local development: Use .env files
OPENAI_API_KEY=sk-...
DATABASE_URL=postgresql://...
Production: Set in deployment configuration
runagent deploy . --env OPENAI_API_KEY=$OPENAI_API_KEY
Secrets are encrypted at rest and never logged.
Default limits:
- Memory: 512MB (configurable up to 8GB)
- Timeout: 30 seconds (configurable up to 900s)
- Payload size: 10MB
- Concurrent requests: Based on plan
Configure in runagent.config.json:
{
"deployment": {
"memory": "2GB",
"timeout": 120
}
}
Deployment Questions
Local deployment:
- Runs on your machine
- Free forever
- Good for development and testing
- Requires your machine to be running
Cloud deployment:
- Runs on RunAgent infrastructure
- Auto-scaling and high availability
- Monitoring and logging included
- Accessible from anywhere
Simply deploy again with the same name:
# Updates existing deployment
runagent deploy . --name my-agent
RunAgent handles:
- Zero-downtime updates
- Automatic rollback on failure
- Version history
Yes! Use different names or environments:
# Production version
runagent deploy . --name agent-prod --env production
# Staging version
runagent deploy . --name agent-staging --env staging
# Development version
runagent deploy . --name agent-dev --env development
SDK Questions
Available now:
- Python
Coming soon:
- JavaScript/TypeScript (Q2 2024)
- Rust (Q2 2024)
- Go (Q3 2024)
You can always use the REST API directly from any language.
Python SDK:
for chunk in client.run_generic_stream(input_data):
print(chunk, end="", flush=True)
Direct API: Use Server-Sent Events (SSE) with appropriate client library.
Yes, the SDK clients are thread-safe. You can:
- Share a client instance across threads
- Use in async contexts with AsyncRunAgentClient
- Handle concurrent requests
Troubleshooting
Common causes:
- Missing environment variables - Check all are set
- File path issues - Use relative paths
- Dependency versions - Pin specific versions
- Resource limits - Increase memory/timeout
Debug with:
runagent logs <deployment-id>
- Add logging to identify slow operations
- Use profiling to find bottlenecks
- Check logs for errors or warnings
- Monitor metrics (coming soon)
import time
start = time.time()
# operation
print(f"Took {time.time() - start}s")
Best Practices
Recommended structure:
my-agent/
├── agent.py # Main agent code
├── utils/ # Helper functions
├── prompts/ # Prompt templates
├── tests/ # Unit tests
├── runagent.config.json
├── requirements.txt
├── .env.example
└── README.md
Use sync when:
- Simple request/response
- No concurrent operations
- Easier to debug
Use async when:
- Multiple API calls
- I/O intensive operations
- Need better performance
Community & Support
- Documentation: docs.run-agent.ai
- Discord: discord.gg/runagent
- GitHub Issues: For bugs and features
- Email: [email protected] for critical issues
Yes! We welcome contributions:
- Code contributions via GitHub
- Documentation improvements
- Bug reports and feature requests
- Community SDKs and templates
See our contributing guide.
The RunAgent CLI and SDKs are open source (MIT license). The deployment infrastructure is proprietary but offers a generous free tier.
Still Have Questions?
Ask on Discord
Get help from the community
Contact Support
For account or billing questions