Agent Configuration

Each agent has an agent.json file that controls its behavior.

agent.json

{
  "name": "Code Assistant",
  "description": "Helps with code review and refactoring",
  "model": {
    "aiResourceId": "uuid-of-ai-resource",
    "modelId": "claude-sonnet-4-6"
  },
  "avatar": "avatar.jpg",
  "parameters": {
    "temperature": 0.7,
    "maxTokens": 4096
  },
  "mcpServers": [],
  "tools": [],
  "rateLimits": {
    "maxRequestsPerMinute": 20
  },
  "guardrails": {
    "maxTurns": 50
  }
}

Fields

model

The AI model to use. References an AI resource configured in Settings > AI.

{
  "aiResourceId": "uuid-of-your-api-key-resource",
  "modelId": "claude-sonnet-4-6"
}

aiResourceId points to an AI resource you’ve set up with your API key. modelId is the model identifier from the provider.

parameters

Field	Type	Default	Description
temperature	number	0.7	Creativity level (0.0 = deterministic, 1.0 = creative)
maxTokens	number	4096	Maximum tokens in the response

rateLimits

Field	Type	Default	Description
maxRequestsPerMinute	number	20	Rate limit for API calls

guardrails

Field	Type	Default	Description
maxTurns	number	50	Maximum conversation turns before warning

avatar

Optional. Path to an avatar image relative to the agent folder (e.g., "avatar.jpg"). Displayed in the chat interface.

mcpServers

Array of MCP server connections. See MCP Servers.

tools

Array of tool identifiers the agent can use. Tools are provided by MCP servers.

File Location

rightplace/agents/{slug}/agents/{agent-name}/agent.json

The agent name is the folder name (e.g., assistant, reviewer, analyst).