Retool AI

Generates complete web/mobile applications by injecting live database schema and permission context into LLM prompts, producing scaffolded apps that respect actual data structure, roles, and access controls without manual template selection. The system introspects connected data sources (Postgres, Databricks, Salesforce) at generation time to ground the LLM output in real schema rather than generic templates, enabling role-based access control and data-level permissions to be enforced from deployment.

Enables users to click on generated app components and @mention data sources or resources to trigger AI-assisted edits that understand full app context, not just isolated component state. Rather than chat-based editing, the system provides the LLM with the complete app structure, current component configuration, and available data sources, allowing edits that maintain consistency across the application and respect existing bindings and permissions.

Enables creation of standalone forms and data collection interfaces with built-in validation, conditional fields, and submission workflows. Forms can be embedded, shared via link, or deployed as standalone apps. Submissions trigger workflows, send notifications, or update databases. Supports file uploads, multi-step forms, and progress tracking.

Generates interactive dashboards and visualizations from database queries and API responses, with support for charts, tables, maps, and custom components. Dashboards update in real-time as underlying data changes and support drill-down, filtering, and export. Visualizations are automatically generated from query results or manually configured.

Enables integration with external APIs through a visual connector interface with support for authentication (API keys, OAuth, basic auth), request/response transformation, error handling, and retry logic. Supports REST, GraphQL, and webhook endpoints. API responses are automatically parsed and can be bound to app components or passed to workflows.

Enables deployment of Retool on-premises or in private cloud environments, maintaining data residency and avoiding cloud data transfer. Self-hosted instances run the full Retool platform (app builder, workflows, agents) with the same features as cloud-hosted deployments. Requires custom annual plan and infrastructure management.

Provides version control for Retool apps with branching, commit history, and release management. Available in Enterprise tier only. Enables teams to collaborate on app development, track changes, and manage releases across environments (staging, production). Integrates with Git or Retool's native version control.

Exposes Retool apps and workflows via REST APIs and webhook endpoints, enabling external systems to trigger workflows, query apps, and integrate with Retool programmatically. Available in Enterprise tier only. Supports authentication, request validation, and response formatting.

Chains AI actions, data transformations, and business logic into executable workflows with built-in conditional branching, exception handling, and full audit trails. Workflows execute on Retool's cloud or self-hosted runtime, with each step logged and monitored, enabling teams to automate complex processes that span multiple data sources and business systems without writing code. Supports scheduled execution, event-triggered execution, and manual triggering with role-based access control.

Deploys custom agents that reason over live data sources and execute tasks autonomously across multiple business systems (CRM, support platforms, databases) with configurable guardrails and permission enforcement. Agents can access multiple data sources, make decisions based on context, and take actions (create records, send messages, generate reports) while respecting user permissions and audit logging. Supports natural language task descriptions and context-aware adaptation based on data state.

Enables users to add documents or data sources to LLM calls with a single click, automatically handling embedding, storage, and retrieval without manual vector database configuration. The system abstracts away RAG complexity (vector storage, embedding model selection, retrieval strategy) and integrates retrieved context directly into AI-powered app generation, workflows, and agent tasks. Mechanism of vector storage and embedding model selection not documented.

Enforces user roles and permissions at both the app and data level, inherited from source systems (Postgres row-level security, Salesforce profiles, etc.) or defined in Retool. Generated apps automatically respect existing permission models, and workflows/agents enforce permissions before executing actions. Supports SSO/SAML integration (Enterprise only) and granular permission controls (Business+ tiers).

Enables embedding Retool apps into external software or websites with authenticated user access and custom branding. Embedded apps inherit Retool's permission model and can be deployed as customer-facing portals or internal tools within third-party applications. Supports user authentication, role-based access, and custom domain configuration.

Enables creation of branded, customer-facing portals with built-in user management, authentication, and role-based access control. Portals are separate from internal apps and support external user provisioning, self-service signup (mechanism not documented), and custom branding. External users are billed separately from internal users at a per-user rate.

Enables creation of reusable component groups (modules) that can be shared across apps and versioned. Free/Team tiers limited to 5 modules; Business/Enterprise tiers support unlimited modules. Modules reduce duplication and enable consistent UI/UX across applications. Version control available in Enterprise tier only.

Provides both SQL and visual query builders for constructing database queries across multiple data sources (Postgres, Databricks, Salesforce, etc.) without writing raw SQL. Queries are bound to app components and automatically re-execute when inputs change. Supports parameterized queries, caching, and query result transformation.