paragon
npx skills add https://docs.useparagon.com
Agent 安装分布
Skill 文档
Capabilities
Paragon enables AI agents and applications to build native product integrations with 130+ SaaS applications through three purpose-built products: ActionKit for real-time synchronous actions, Managed Sync for data ingestion pipelines, and Workflows for asynchronous event-driven automation. Agents can read and write data across CRMs, file storage, ticketing systems, and other business tools while Paragon handles authentication, data normalization, permissions, and infrastructure.
Skills
ActionKit – Real-Time Integration Actions
AI Agent Tool Calling
- Access hundreds of pre-built integration actions as function tools via API or MCP server
- Execute synchronous (real-time) operations with extremely low-latency SLAs
- Use agent-optimized JSON Schema specs for accurate tool calling
- Examples: Create Jira tasks from meeting transcripts, query Shopify sales reports, generate Google Docs drafts
API Endpoints
- Base URL:
https://actionkit.useparagon.com/projects/[Project ID] - Authentication: Bearer token with Paragon User Token (JWT)
- List available actions:
GET /projects/[Project ID]/actions - Execute action:
POST /projects/[Project ID]/actionswith action name and parameters
Universal Interface Features
- Consistent filter schema across all list/search operations (contains, equals, greater_than, etc.)
- Dynamic field reloading at query-time for custom fields in third-party apps
- Support for 100+ integrations including Salesforce, HubSpot, Jira, Slack, Stripe, Shopify, and more
Implementation Examples
- Vercel AI SDK integration with tool definitions
- LangChain/LangGraph implementation for agentic workflows
- OpenAI client library with function calling
- Custom implementations with any LLM supporting tool/function calling
Managed Sync – Data Ingestion Pipelines
Sync API
- Base URL:
https://sync.useparagon.com - Enable syncs:
POST /syncsto start data ingestion from user’s integration - Get sync status:
GET /syncs/[sync_id]to check ingestion progress - Pull synced records:
GET /syncs/[sync_id]/recordswith pagination support - Download content:
GET /syncs/[sync_id]/records/[record_id]/contentfor file data
Normalized Data Objects
- File schema: Files, folders, permissions from cloud storage (Google Drive, Dropbox, OneDrive, Box, Confluence, SharePoint)
- Contact schema: Contacts, companies, deals from CRMs (Salesforce, HubSpot, Pipedrive, Zoho)
- Company schema: Organization data with custom fields
- Deal schema: Sales opportunities and pipeline data
- Ticket schema: Support tickets from Zendesk, Freshdesk, ServiceNow
- Custom object schema: User-defined data structures
Permissions API
- Check access:
POST /permissions/check-accessto verify user access to records - Batch check:
POST /permissions/batch-check-accessfor multiple records - List users:
GET /permissions/usersto enumerate users with access - List objects:
GET /permissions/objectsto find accessible records - Expand relationships:
POST /permissions/expandto traverse permission graph
Sync Lifecycle
- Initial sync: Full data pull and indexing when sync is enabled
- Incremental syncs: Periodic checks for changes (every minute)
- Periodic full refresh: 24-hour full sync to detect deletions and drift
- Webhook notifications:
sync_completed,record_created,record_updated,record_deletedevents
Workflows – Event-Driven Automation
Workflow Builder (Visual)
- Drag-and-drop interface for building multi-step automations
- No-code approach for non-technical teams
- Deploy workflows to customers via Connect Portal
Paragraph (Code-First)
- TypeScript framework for programmatic workflow definition
- Full control and flexibility with npm library support
- Version control integration with Git sync
- Seamless switching between UI and code representations
Workflow Triggers
- App events: Triggered by events in your application
- Integration triggers: Fired by webhooks from third-party apps
- Integration-enabled triggers: When user enables an integration
- Scheduler triggers: Time-based execution (cron-like)
- API resource triggers: Custom webhook endpoints
- Request triggers: HTTP POST webhooks
Workflow Steps
- Integration actions: Perform operations in connected apps (send Slack message, create Salesforce record)
- API requests: Call any external API with dynamic data
- Functions: Custom JavaScript with npm library support
- Conditionals: Branch logic based on data
- Fan-out: Parallel iteration over lists with result reduction
- Dynamic variables: Reference data from upstream steps with
{{syntax
Workflow Features
- Durable job infrastructure with queues and dynamic scaling
- Auto-retry configuration for resilience
- Rate limit handling for third-party APIs
- File handling and transformation
- Version history and rollback
- Environment secrets management
Connect Portal – Authentication Layer
Embedded SDK
- React component for seamless in-app integration experiences
- White-labeled customization options
- Multi-tenant architecture with isolated customer data
Headless Connect Portal
- Full control over UI while leveraging Paragon’s authentication backend
- Custom integration experience tailored to product needs
- Same security and authentication handling as standard portal
Features
- Fully-managed OAuth 2.0 and API key authentication
- Automatic credential refresh and secure storage (encrypted at rest and in-transit)
- User settings and field mapping configuration
- Workflow visibility and permissions management
- Custom dropdown fields for user configuration
Proxy API – Direct Integration Access
Base URL
- Cloud:
https://proxy.useparagon.com - On-premise:
https://worker-proxy.[your-host]
Request Format
- Path:
/projects/[Project ID]/sdk/proxy/[Integration Type]/[API Path] - Custom integrations:
/projects/[Project ID]/sdk/proxy/custom/[Integration ID]/[API Path] - Authentication: Bearer token with Paragon User Token
- Supports any HTTP verb (GET, POST, PUT, DELETE, etc.)
Use Cases
- One-off API requests (fetch Salesforce contacts)
- Access API methods not available in Workflow actions
- Complex custom integration logic
- Migrating existing integration code
Special Features
- Raw response handling:
X-Paragon-Use-Raw-Responseheader for binary data - Custom base URL support: Use fully-qualified URLs to override default endpoints
- File download support: Retrieve binary content like PDFs and images
Platform APIs
User Management API
- Manage connected users and their integration credentials
- Track user authentication status and integration connections
Task History API
- View detailed logs of all integration executions
- Monitor workflow runs and action executions
- Track errors and retry attempts
Multi-Account Authorization
- Connect multiple accounts of same integration type per user
- Specify credential with
X-Paragon-Credentialheader - Support for complex multi-account scenarios
Workflows
Implementing RAG with File Ingestion
- Enable Managed Sync for user’s file storage integration (Google Drive, Dropbox, etc.)
- Wait for
sync_completedwebhook or poll sync status - Paginate through synced files using Sync API
- Check user permissions with Permissions API to filter accessible files
- Download file content using
GET /syncs/[sync_id]/records/[record_id]/content - Ingest normalized file data into RAG pipeline
- Handle incremental updates via
record_createdandrecord_updatedwebhooks
Building Bidirectional CRM Sync
Pull Direction (Managed Sync)
- Enable Managed Sync for CRM (Salesforce, HubSpot, Pipedrive)
- Receive webhook notifications for contact/deal changes
- Pull updated records via Sync API
- Store in application database
Push Direction (Workflows)
- Create workflow triggered by app events (new contact created)
- Add integration action step to create/update record in CRM
- Map application fields to CRM fields
- Deploy workflow to customers
- Monitor execution in Task History
AI Agent Tool Integration
- Fetch available actions:
GET /actionkit.useparagon.com/projects/[Project ID]/actions - Convert JSON Schema specs to LLM tool definitions
- Pass tools to LLM with agent-optimized descriptions
- When agent selects tool, execute:
POST /actionkit.useparagon.com/projects/[Project ID]/actions - Return results to agent for continued conversation
- Handle errors and retry failed actions
Custom Integration Workflow
- Create custom integration in Paragon dashboard
- Define API endpoints and authentication method
- Build workflow using custom integration actions
- Test with preview user in Connect Portal
- Deploy to production customers
- Monitor via Task History and event logs
Integration
Third-Party Integrations Supported
- CRM: Salesforce, HubSpot, Pipedrive, Zoho, Microsoft Dynamics, Outreach, Close
- File Storage: Google Drive, Dropbox, OneDrive, Box, Amazon S3, Confluence, SharePoint
- Ticketing: Zendesk, Freshdesk, ServiceNow, ClickUp, Monday.com
- Communication: Slack, Microsoft Teams, Gmail, Outlook
- Project Management: Jira, Linear, Asana, GitHub
- Finance: Stripe, QuickBooks, NetSuite, Xero, Sage Intacct
- Marketing: Mailchimp, Klaviyo, Marketo, Pardot
- HR: BambooHR, Gusto, Workday
- Custom integrations: Build your own with webhook support
Deployment Options
- Cloud (US and EU regions)
- On-premise managed deployments
- Forward-deployed options
- Flexible infrastructure choices
Monitoring & Observability
- Task History: Detailed logs for every integration execution
- Event Destinations: Send logs to Datadog, New Relic, Sentry, Slack
- Event logs: Track all integration activity
- Error tracking and retry monitoring
Context
Authentication & Security
- Paragon handles all OAuth 2.0 flows and API key management
- Credentials encrypted at rest and in-transit
- Multi-tenant isolation ensures customer data separation
- JWT-based user tokens for API authentication
- Support for multi-account authorization per user
Data Normalization
- Managed Sync normalizes data from different sources into consistent schemas
- Custom fields from third-party apps are automatically included
- Permissions graph built from integration RBAC models
- Consistent interface across 130+ integrations
Performance & Reliability
- Real-time SLAs for ActionKit (synchronous operations)
- Durable job infrastructure for Workflows with automatic retries
- Intelligent backoff and retry scheduling for syncs
- Rate limit handling and pagination support
- High-availability runtime with dynamic scaling
Development Flexibility
- Choose between visual (Workflow Builder) or code-first (Paragraph) approaches
- Switch between UI and code representations seamlessly
- Version control integration for workflows
- Support for custom JavaScript and npm libraries
- Headless options for custom UI implementations
Billing & Limits
- Connected users: Tracked per unique user with integrations
- Tasks: Counted per workflow execution or API call
- Concurrency limits: Configurable based on plan
- Sync volume: Managed automatically with intelligent scheduling
For additional documentation and navigation, see: https://docs.useparagon.com/llms.txt