> ## Documentation Index > Fetch the complete documentation index at: https://docs.flowra.dev/llms.txt > Use this file to discover all available pages before exploring further. # Create workflow (agent) > Create a new agent/workflow with name, tools, system prompt, and model. Can be triggered manually or by schedules/webhooks. ## OpenAPI ````yaml /api-reference/openapi.json post /api/v1/workflow/manager/create openapi: 3.0.0 info: title: Flowra API (API Key Endpoints) description: >- **Flowra** is an AI-powered platform for building agents, workflows, and integrations. ### Capabilities - **Tools & Toolkits** – Define and group tools for agents and workflows. - **Workflows** – Orchestrate steps and automate processes. - **Triggers** – Schedule or event-based execution. - **Auth & Connections** – OAuth, API keys, and connected accounts. - **MCP** – Model Context Protocol server integration. - **Knowledge / RAG** – Add files and URLs for retrieval-augmented generation. ### Authentication Use the **x-api-key** header with a project API key to access endpoints that support API Key auth. Create and manage API keys in the dashboard under your project settings. Only endpoints that support API Key authentication via the @ApiKeyEnabled() decorator. Send x-api-key (required). On multi-tenant endpoints, optionally send x-username to act as an external user. Only endpoints that support API Key authentication via the @ApiKeyEnabled() decorator. Send x-api-key (required). On multi-tenant endpoints, optionally send x-username (defaults to the project’s default external user). version: '1.0' contact: name: Flowra url: https://flowra.dev servers: - url: https://flowra.dev description: Flowra API security: [] tags: - name: Users description: User profile - name: External Users description: End-user identities inside a project (x-username) - name: Auth Configs description: OAuth and API key configs for toolkits - name: Connected Accounts description: OAuth connected accounts and link creation - name: Toolkits description: List and search toolkits - name: Tools description: List, execute, and manage tools - name: Skills description: Browse, create, import, and install skills - name: Workflow & Agent description: Create, run, and inspect workflows and agents - name: Triggers description: Schedule and event-based trigger instances - name: Chat description: Threads and streaming agent runs - name: File description: Upload, list, and download files - name: Knowledge description: RAG knowledge collections - name: Table description: Collections and documents - name: Sandbox description: Ephemeral and durable code sandboxes - name: Browser description: Live browser sessions and view - name: LLM description: List AI models and generate completions - name: MCP description: MCP server management paths: /api/v1/workflow/manager/create: post: tags: - Workflow & Agent summary: Create workflow (agent) description: >- Create a new agent/workflow with name, tools, system prompt, and model. Can be triggered manually or by schedules/webhooks. operationId: WorkflowManagerController_create parameters: - name: x-username in: header required: false description: >- External user (end user) to act as. Optional — if omitted, Flowra uses the project’s default external user. Send a stable username (e.g. customer_alice) only when you need another end-user context. schema: type: string default: (project default user) example: customer_alice requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/CreateAgentDto' responses: '200': description: Agent created successfully content: application/json: schema: $ref: '#/components/schemas/WorkflowResponseDto' '201': description: '' security: - x-api-key: [] - x-api-key: [] x-username: [] components: schemas: CreateAgentDto: $ref: '#/components/schemas/CreateAgentDto' type: object properties: name: type: string description: Agent name default: New Agent description: type: string description: Agent description agentPrompt: type: string description: System prompt / instructions for the agent modelSettings: description: Model and generation settings allOf: - $ref: '#/components/schemas/CreateAgentModelSettingsDto' toolSlugs: description: Tool slugs to attach to the agent type: array items: type: string allowedKnowledgeCollectionIds: description: >- Knowledge collection IDs this agent can access (from RAG_LIST_COLLECTIONS). If set, RAG tools are restricted to these collections only. type: array items: type: string allowedDatabaseCollectionIds: description: >- Database collection IDs (from DATABASE_LIST_COLLECTIONS) this agent can access. If set, DATABASE_* tools are restricted to these collections only. type: array items: type: string databaseAccessPolicies: type: object description: >- Per-collection end-user access policies (keyed by collectionId). Controls which operations, row scope (own/all), and columns end users may use through this agent. additionalProperties: true isPublic: type: boolean description: >- Whether this agent is public (visible in other projects’ agent select) default: false externalMcpConfigs: description: >- External MCP configs: [{ name, url, headers?, tools }]. tools = selected tool slugs from that MCP. type: array items: $ref: '#/components/schemas/CreateAgentExternalMcpConfigDto' agentCapabilities: description: >- User-facing agent capabilities (advanced planning, summarization, retry, approval, etc.). These are mapped server-side to runtime selection and middleware. allOf: - $ref: '#/components/schemas/CreateAgentCapabilitiesDto' toolsMode: enum: - NORMAL - SMART type: string description: >- Agent tools: NORMAL = explicit list, SMART = meta + discovery filters (same as MCP owner toolsMode). discoveryToolsWhitelist: type: boolean description: >- When toolsMode is SMART: true = discovery limited to pinned toolkit slugs; false = exclude those slugs from discovery. subAgentIds: description: >- Child agent workflow IDs for multi-agent coordination (one level). Platform enables deeper planning automatically. type: array items: type: string skillIds: description: >- Skill IDs to attach as playbooks. Platform enables deeper planning automatically. Skills do not merge tools into toolSlugs. type: array items: type: string starterPrompts: description: >- Suggested starter prompts shown in chat empty state (max 8, each max 200 chars). type: array items: type: string maxLength: 200 avatarUrl: type: string maxLength: 500 description: >- Optional agent avatar. Use `file:` from file upload (toolkit_logo / agent_avatar) or an https image URL. example: file:019f0000-0000-7000-8000-000000000001 triggers: description: >- Optional cron / channel event triggers (0..N). Dashboard chat is always available and must not be listed. kind must be cron or trigger_tool. type: array items: $ref: '#/components/schemas/CreateAgentTriggerDto' outputSchema: type: object description: >- Optional structured shape for the agent final answer (API / integrations). Map of field name → { type, required?, description? }. connectionCallbackUrl: type: string maxLength: 2048 WorkflowResponseDto: $ref: '#/components/schemas/WorkflowResponseDto' type: object properties: id: type: string name: type: string description: type: string projectId: type: string nodes: type: array items: $ref: '#/components/schemas/WorkflowNodeResponseDto' edges: type: array items: $ref: '#/components/schemas/WorkflowEdgeResponseDto' triggers: type: array items: $ref: '#/components/schemas/WorkflowTriggerResponseDto' workflowType: type: string enum: - static - agent inputSchema: type: object outputSchema: type: object isActive: type: boolean isPublic: type: boolean description: Whether this workflow (agent) is public for select in other projects version: type: string description: >- Version in form 1.0.x; incremented on each update (for marketplace update check) clonedFromWorkflowId: type: string nullable: true description: 'If cloned from marketplace: source workflow id' clonedFromVersion: type: string nullable: true description: Source workflow version at clone time (e.g. 1.0.1) executionCount: type: number successfulExecutionCount: type: number failedExecutionCount: type: number lastExecutedAt: type: string metadata: $ref: '#/components/schemas/WorkflowMetadata' connections: type: array items: $ref: '#/components/schemas/WorkflowConnectionDto' tools: type: array items: $ref: '#/components/schemas/WorkflowToolDetailDto' createdAt: format: date-time type: string updatedAt: format: date-time type: string sourceVersionLatest: type: string example: 1.0.2 toolsMode: type: string enum: - NORMAL - SMART discoveryToolsWhitelist: type: boolean connectionCallbackUrl: type: string nullable: true required: - id - name - projectId - nodes - edges - triggers - workflowType - isActive - isPublic - version - executionCount - successfulExecutionCount - failedExecutionCount - connections - tools - createdAt - updatedAt securitySchemes: x-api-key: type: apiKey in: header name: x-api-key description: >- Project API key. Create and manage keys in the dashboard under Project → API Keys. Send as header: x-api-key: x-username: type: apiKey in: header name: x-username description: >- Optional. External user username. If omitted, Flowra uses the project’s default external user. Send as header: x-username: ````