Endpoint
Authentication
Requires JWT authentication viaAuthorization: Bearer <token> header.
Description
Creates a new AI agent with a custom name, icon, and initial configuration version. The agent is initialized with default tools and system prompt, which can be customized immediately or later via the update endpoint.Limits
Agent creation is subject to tier-based limits:- Free tier: 3 agents maximum
- Pro tier: 10 agents maximum
- Enterprise tier: Unlimited agents
402 error with code AGENT_LIMIT_EXCEEDED.
Request Body
Agent name (2-100 characters). Used for identification in the UI.
Whether this agent should be the default for new threads. Only one agent can be default per account.
Icon identifier from the icon library (e.g., “bot”, “code”, “brain”, “sparkles”).
Icon color in hex format (e.g., “#6366F1”).
Icon background color in hex format.
Custom system prompt for the agent. If not provided, uses default Suna configuration.
Array of configured MCP (Model Context Protocol) integrations.
Array of custom MCP server configurations.
Tool configurations. Core tools (bash, read, write, edit, glob, grep) are automatically enabled.
Response
Returns the created agent with full configuration including the initial version (v1).Unique identifier for the agent (UUID format)
Agent name
Agent description
Icon identifier
Icon color in hex format
Icon background color in hex format
Whether this is the default agent
ID of the current active version
Total number of versions (starts at 1)
Full version configuration object
ISO 8601 timestamp of creation
ISO 8601 timestamp of last update
Examples
Response Example
Error Responses
Agent limit exceeded for current tier
Invalid request data (e.g., missing name, invalid icon)
Missing or invalid authentication token
Server error during agent creation
Notes
- Default Agent: Setting
is_default: trueautomatically unsets any existing default agent - Versioning: Every agent starts with version v1, which is automatically created
- Core Tools: The core tools (bash, read, write, edit, glob, grep) are always enabled and cannot be disabled
- Icon Library: Use the
/agents/generate-iconendpoint to get AI-suggested icons based on agent name - Cache: Creating an agent invalidates the agent count cache for the user
Related Endpoints
- Update Agent - Modify agent configuration
- Get Agent - Retrieve agent details
- Delete Agent - Remove an agent
- Run Agent - Start an agent run