AmaStudio Agent Runtime

AmaStudio's agent runtime manages AI agent sessions, messages, and tasks. The agent runtime API stores agent data on amadev.org so it persists across devices.

Architecture

Agent execution stays local in AmaStudio. Only the data (config, sessions, messages, tasks) is synced to amadev.org.

Code

┌─────────────────────────────────┐
│         AmaStudio               │
│  ┌──────────┐  ┌─────────────┐ │
│  │ Execution│  │  Data Sync  │ │
│  │  (Local) │──│ → amadev.org│ │
│  └──────────┘  └─────────────┘ │
└─────────────────────────────────┘

Database Schema

The agent runtime uses these Prisma models:

Agent — Agent configuration (model, instructions, tools)
AgentSession — A conversation or task session
AgentMessage — Messages within a session
AgentTask — Tasks assigned to an agent

API Endpoints

Agents

Code

GET /v1/agents                    # List all agents
POST /v1/agents                   # Create agent
GET /v1/agents/:id                # Get agent details
PUT /v1/agents/:id                # Update agent config
DELETE /v1/agents/:id             # Delete agent

Sessions

Code

GET /v1/agents/sessions            # List sessions
POST /v1/agents/sessions           # Create session
GET /v1/agents/sessions/:id        # Get session details

Messages

Code

GET /v1/agents/sessions/:id/messages    # List messages
POST /v1/agents/sessions/:id/messages   # Send message

Tasks

Code

GET /v1/agents/tasks               # List tasks
POST /v1/agents/tasks              # Create task
PUT /v1/agents/tasks/:id           # Update task status

Sync from AmaStudio

AmaStudio's AmadevService provides IPC methods for agent sync:

AmadevSyncAgents — triggers agent config sync
Agent sessions and messages are persisted automatically
Data is available across all connected AmaStudio instances

Data Ownership

All agent data is scoped to the authenticated user. Each API key is tied to a user, and data access is limited to that user's API key.