Supported Frameworks & Deployment Types
When registering an agent, you need to select the framework (how it was built) and the deployment type (where it runs). This page explains each option in detail to help you choose correctly.
Frameworks
A framework is the technology your agent was built with. TraptureIQ uses this information to communicate with your agent using the correct API protocol.
ADK is the only supported framework at this time. Support for LangGraph and CrewAI is planned for upcoming phases.
| Framework | What It Is | When to Choose | Availability |
|---|---|---|---|
| ADK | Google's Agent Development Kit for building AI agents | Your agent was built using the ADK SDK | ✅ Available |
| LangGraph | LangChain's orchestration framework for stateful, multi-step agents | Your agent uses the LangGraph library for orchestration | 🔜 Coming soon |
| CrewAI | Framework for building teams of specialized AI agents that collaborate | Your agent uses CrewAI for multi-agent workflows | 🔜 Coming soon |
How to Know Which Framework to Choose
If you're unsure which framework your agent uses:
- Check your agent's codebase — look for imports like
google.adk,langgraph,crewai, etc. - Check your agent's API endpoint — try sending a request and see what format it responds in
- Ask the developer who built the agent
- Currently, only ADK agents can be registered — support for other frameworks is coming soon
Deployment Types
The deployment type tells TraptureIQ where your agent is hosted and how to connect to it.
| Type | What It Is | Authentication Required? | Best For |
|---|---|---|---|
| Agent Engine | Vertex AI Agent Engine — a Google-managed hosting service for AI agents | Yes (one of three auth methods) | Agents deployed through Vertex AI's managed infrastructure |
| Cloud Run | Google Cloud Run — containerized services that scale automatically | No | Agents deployed as containers on Google Cloud |
| Custom | Any other server — on-premises, another cloud provider, local development machine | No | Agents hosted anywhere outside of GCP managed services |
How to Know Which Deployment Type to Choose
| Where Your Agent Is Hosted | Choose |
|---|---|
| Vertex AI Reasoning Engine | Agent Engine |
| Google Cloud Run | Cloud Run |
| AWS, Azure, or another cloud | Custom |
| Your own on-premises servers | Custom |
| Your local development machine | Custom |
Authentication Methods (Agent Engine Only)
When deploying on Agent Engine, TraptureIQ needs to authenticate to your GCP project. Three options are available:
| Method | How It Works | Security Level | Setup Time |
|---|---|---|---|
| Authorize with Impersonation | TraptureIQ temporarily impersonates a service account you create. No credentials are exchanged — only short-lived tokens. | Highest | ~5 min |
| Authorize TraptureIQ Principal | You grant TraptureIQ's own service account direct Google Cloud permissions on your project | High | ~2 min |
| Upload SA Key | You upload a service account key JSON file, which TraptureIQ encrypts and stores | Standard | ~5 min |
For detailed setup instructions, see the Authentication Setup guide.
Cloud Run and Custom deployments do not require authentication setup — TraptureIQ connects directly over HTTPS.
Framework + Deployment Combinations
Not all combinations are equally common. Here's what we typically see:
| Framework | Common Deployment | Notes |
|---|---|---|
| ADK | Agent Engine, Cloud Run | ADK agents are most commonly deployed on Agent Engine or Cloud Run |
| LangGraph | Cloud Run, Custom | LangGraph agents are typically deployed as containers on Cloud Run or custom servers |
| CrewAI | Cloud Run, Custom | CrewAI agents are typically containerized and deployed on Cloud Run |
Tips for Beginners
- If the Test Connection fails with "framework handshake error" — you may have selected the wrong framework. Try a different one.
- LangGraph and CrewAI support is coming in a future phase — for now, only ADK agents can be registered.
- Cloud Run is the easiest deployment type — no authentication setup needed; just provide the HTTPS URL.
- You can change the framework later — if you chose wrong, edit the agent and select a different framework.