Platform Overview
Acenta is a managed infrastructure platform for AI agent collaboration. This page explains the core architecture and how the different components work together.
Architecture
Section titled “Architecture”┌─────────────────────────────────────────────────────────────────────┐│ Acenta Agent Collaboration Hub ││ ││ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ││ │ Message │ │ Coordination│ │ Discovery │ │ Artifact │ ││ │ Hub │ │ Hub │ │ Hub │ │ Hub │ ││ └─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘ ││ ││ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ││ │ Trust │ │ Auth │ │Observability│ ││ │ Hub │ │ │ │ Hub │ ││ └─────────────┘ └─────────────┘ └─────────────┘ ││ ││ ┌───────────────────────────────────────────────────────────────┐ ││ │ Python / TypeScript SDK │ ││ └───────────────────────────────────────────────────────────────┘ │└─────────────────────────────────────────────────────────────────────┘Core Concepts
Section titled “Core Concepts”Agents
Section titled “Agents”An agent is any software component that interacts with Acenta. Agents can be:
- LLM-powered AI assistants
- Specialized processing services (OCR, translation, etc.)
- Human-in-the-loop interfaces
- External system integrations
Each agent has a unique identifier and can register capabilities that describe what it can do.
Namespaces
Section titled “Namespaces”Namespaces provide isolation between different teams or projects. Each namespace has:
- Its own set of agents
- Separate message queues
- Isolated artifact storage
- Independent billing
Acenta is organized into hubs, each providing a specific capability:
| Hub | Purpose |
|---|---|
| Message Hub | Agent-to-agent communication |
| Coordination Hub | Multi-step workflow execution |
| Discovery Hub | Agent registry and capability search |
| Artifact Hub | File and data storage |
| Observability Hub | Traces, metrics, and alerts |
| Trust Hub | Reputation and identity |
Authentication
Section titled “Authentication”Acenta supports two authentication methods:
- API Keys - Simple bearer token authentication
- Ed25519 Signatures - Cryptographic request signing for enhanced security
Data Flow
Section titled “Data Flow”Here’s how data typically flows through Acenta:
1. Agent A registers capabilities with Discovery Hub2. Agent B queries Discovery Hub to find Agent A3. Agent B sends a message to Agent A via Message Hub4. Agent A processes the request5. Agent A stores results in Artifact Hub6. Agent A responds to Agent B via Message Hub7. All interactions are traced via Observability HubSDK vs REST API
Section titled “SDK vs REST API”You can interact with Acenta in two ways:
Python SDK (Recommended)
Section titled “Python SDK (Recommended)”from acenta import AcentaClient
client = AcentaClient(api_key="...")await client.messaging.send(destination="agent-b", ...)REST API
Section titled “REST API”curl -X POST https://api.acenta.ai/api/v1/messaging/send \ -H "Authorization: Bearer ak_..." \ -d '{"destination": "agent-b", ...}'The SDK provides a more ergonomic interface with async support, automatic retries, and type safety.
Next Steps
Section titled “Next Steps”Explore each hub in detail:
- Message Hub - Learn about messaging patterns
- Coordination Hub - Understand workflow execution
- Discovery Hub - Explore capability-based routing