Agents
Configure agents with tools, knowledge bases, workflows, and LLM models
Capabilities
- Tools — Call APIs, query databases, send notifications
- Knowledge Bases — Semantic search over documents (RAG)
- Workflows — Execute multi-step deterministic processes
- Context — Maintain conversation history across messages
- Rich Widgets — Return structured data rendered as interactive UI components (charts, tables, forms). Define JSON schemas for your widgets, and the agent outputs data that the client renders.
LLM Models
Ensemble supports models from major providers.
| Provider | Fast | Standard | Reasoning |
|---|---|---|---|
| Anthropic | Claude Haiku 4.5 | Claude 3.7 Sonnet Claude Sonnet 4 Claude Sonnet 4.5 Claude Sonnet 4.6 | Claude Opus 4.6 |
| OpenAI | GPT-4.1 Mini GPT-5 Mini | GPT-4.1 GPT-5 GPT-5.2 | o3 |
| Gemini 2.5 Flash | Gemini 2.5 Pro Gemini 3 Pro Gemini 3.1 Pro | - | |
| Vertex AI | Claude Haiku 4.5 | Claude Sonnet 4.5 | - |
| DeepSeek | - | DeepSeek V3.2 | - |
| MoonshotAI | - | Kimi K2.5 | - |
| Zhipu | - | GLM-5 | - |
| Cerebras | GPT-OSS 120B | - | - |
Bring Your Own Model
Tenants can add custom models that aren't in the default list. In the admin settings, configure:
- API Endpoint — The model's API URL (must be OpenAI-compatible)
- API Key — Authentication key for the endpoint
- Model Name — Display name in the model selector
- Model Family — Category for cost tracking and defaults
Custom models work with any OpenAI-compatible API, including self-hosted models (vLLM, Ollama, etc.) and other cloud providers.
Multi-Agent Orchestration
Every Ensemble agent is an orchestration agent. Beyond calling tools, executing workflow and accessing knowledge bases, agents can delegate complex tasks to sub-agents — enabling sophisticated multi-agent workflows out of the box.
Default Agent
Handles requests using its own tools and knowledge bases, and can delegate specific tasks to sub-agents when configured. Each sub-agent runs in its own context window and returns results to the parent for synthesis.
Best for: Single-domain tasks where one agent does most of the work, delegating complex tasks to sub-agents as needed.
Similar to: Claude Code
Supervisor Agent
A coordinator that routes requests to specialized sub-agents without handling tasks directly. The supervisor determines which sub-agent handles each request based on the task. Sub-agents share context with the supervisor but execute independently.
Best for: Multi-domain applications where each sub-agent specializes in a different area (e.g., billing, support, inventory).
Similar to: Langgraph Supervisor
Versioning
Agents support Versioning.