Create System Integrations Through Conversation
Connectors are the core of FIM One. A connector represents an external system's interface — Agent reads and writes data through connectors, and the platform auto-handles authentication and request management.
Three Ways to Connect Any System
FIM One's connectors don't depend on pre-built integrations. Your team can quickly create them through any of these methods:
Import OpenAPI Docs
Upload the target system's Swagger / OpenAPI file (YAML, JSON, or URL), auto-parse and generate complete connectors with all Actions. Create new or append to existing.
AI Conversation Generation
Describe interface needs in natural language in the AI panel, or provide API docs for AI to interpret. Supports conversational iteration — generate → modify → test → publish, all in conversation.
MCP Ecosystem
If the target system already has a community-maintained MCP Server, search and one-click install from MCP Hub. You can also develop custom MCP Servers.
Import API Docs, Auto-Generate Connectors
Import OpenAPI spec files (YAML, JSON, or URL), and the system auto-parses interface definitions, generating complete connectors with all Actions in one go.
"Supports two modes: create new (generate a brand new connector from docs) and append (add new Actions to existing connectors). Each generated Action automatically includes method, path, parameter definitions, and LLM-readable tool descriptions, ready for Agent to recognize and invoke."
Create Connectors via AI Conversation
No standard API docs? No problem. In the Action editor's built-in AI panel, describe your interface needs in natural language:
Database Connectors
Direct SQL access to PostgreSQL, MySQL, Oracle, SQL Server — and the enterprise databases deployed widely in China (DM, KingbaseES, GBase, Highgo) that most global AI platforms cannot reach. Schema introspection, AI-powered annotation, read-only query execution, and encrypted credentials at rest.
Connector Progressive Disclosure
A single ConnectorMetaTool replaces per-action tools. The system prompt receives lightweight stubs only (name + one-line description, ~30 tokens per connector vs ~250 tokens per action). Agent calls discover(connector) to load full action schema on demand.
Semantic Schema Annotations
Extend connector schema fields with semantic_tag, description, and pii flags. Annotations are surfaced in LLM tool descriptions, enabling the model to understand field semantics and handle sensitive data appropriately.
Import / Export / Fork
Share connector templates as JSON files. Clone and customize existing connectors to quickly bootstrap new integrations without starting from scratch.
Automatic Credential Handling
Connectors support multiple authentication methods, automatically injected at runtime:
Agent doesn't need to handle auth details when calling connectors — the platform auto-attaches credentials before requests are sent.
Enterprise
Need private deployment, custom connectors, or professional support? Our team is ready to help you scale your AI transformation.