> ## Documentation Index > Fetch the complete documentation index at: https://docs.prisme.ai/llms.txt > Use this file to discover all available pages before exploring further. # Connectors > Connect Prisme.ai to external services via App import or MCP tools for AI agents Connectors provide ready-to-use integrations with popular external services. Each connector can be used in two ways depending on your use case. ## Integration Modes Import the connector as an app into your workspace and call automations directly from DSUL. **Best for:** Workflow automations, data pipelines, scheduled jobs Connect the MCP endpoint to Agent Factory agents. They discover and use tools automatically. **Best for:** AI agents, conversational interfaces, dynamic tool selection ## Available Connectors Extract text and structured data from images and documents with Azure Computer Vision and Document Intelligence, from Agent Factory agents (MCP) or Builder workflows. Search the web from AI agents through Bing Grounding in Azure AI Foundry. Web search and LLM-ready grounding context via Brave Search API. Exposes `brave_search` (raw results) and `brave_llm_context` (structured snippets) as MCP tools. Read and write Confluence Cloud pages, spaces, comments, labels and attachments via eight MCP entity tools, with CQL search and per-user Atlassian OAuth (tenant-context App+MCP). Read/write access to the DataGalaxy data catalog: sources, structures, glossary, links, comments and tasks. Read and write Excel workbooks stored in OneDrive or SharePoint. 50+ operations for worksheets, ranges, tables, and charts. Read Figma files, comments, components, styles and library data. 13 entity tools covering files, images, webhooks, variables and analytics. Manage GitLab projects, issues, merge requests and CI/CD pipelines. 16 entity tools covering projects, MRs, pipelines, branches, commits, releases and more. Bridge a Google Chat bot to a Prisme.ai AI Knowledge agent through an HTTPS webhook. Per-thread context, async acknowledgement and service-account-signed replies. Run Google web searches from AI agents — synthesized grounded answers (Gemini) or rich filtered results and image search (Custom Search), via a Google API key or service account. Drive, Docs, Sheets, Gmail and Calendar through five MCP entity tools, with per-user OAuth (central or tenant client), service-account JWT or a direct access token. Manage time tracking in Gryzzly: customers, projects, tasks, declarations, leave periods and payroll exports. Read and write HubSpot CRM and Marketing data through MCP with per-user OAuth (central or tenant client) or a Private App token — \~120 operations across contacts, companies, deals, tickets, engagements, lists, forms, emails and workflows. Search, create and manage Jira Cloud issues, projects, users, metadata and agile boards via seven MCP entity tools, with per-user Atlassian OAuth (tenant-context App+MCP). Legacy Jira connector (static Basic-Auth / PAT credential, flat tool names, Cloud and Data Center). Superseded by the new Jira connector. Manage monday.com boards, items, columns, docs and users via the v2 GraphQL API. Outlook mail, calendar and contacts via Microsoft Graph, behind six MCP entity tools, with per-user Microsoft Entra OAuth (tenant-context App+MCP). Legacy Outlook mail connector (Microsoft Graph, application or delegated-OAuth via mcp-api-key). Superseded by the new Outlook connector. Query Power BI datasets with DAX and manage workspaces, reports, dashboards, refreshes and admin APIs through eleven MCP entity tools, with per-user Microsoft Entra ID OAuth. Wraps the Prisme.ai Storage backend (files, vector stores, crawling, RBAC, API keys, skills). 38 operations, forward-auth — no per-tenant API key. Read/write Salesforce CRM data via REST API v62.0: records, SOQL/SOSL, Bulk API 2.0, Composite, Tooling, Process & Approvals, Reports and Metadata deploys. Query and manage SAP LeanIX Enterprise Architecture — fact sheets (Pathfinder GraphQL), Integration API (LDIF sync) and surveys, via a LeanIX API token. Manage ServiceNow ITSM tickets, change requests, problems and service catalog with Basic or OAuth2 auth. SharePoint sites, document libraries, files, lists and permissions via Microsoft Graph, behind seven MCP entity tools, with per-user Microsoft Entra OAuth (tenant-context App+MCP). Legacy SharePoint connector (Microsoft Graph, application or delegated-OAuth). Superseded by the new SharePoint connector. Read and act on SonarQube / SonarCloud code quality data: projects, issues, hotspots, measures, quality gates, rules and badges. Read and act on Tableau Cloud / Server: REST API 3.x, Metadata GraphQL, VizQL Data Service and Pulse from DSUL or AI agents. Browse, read and write files on any WebDAV server (Nextcloud, ownCloud, generic HTTPS WebDAV). 10 file/directory operations with Basic or Bearer auth. Build, upload, parse, convert, version, and manage Microsoft Word documents stored in OneDrive or SharePoint. ## Choosing the Right Mode | Criteria | App Mode | MCP Mode | | ------------------ | ------------------------------------- | ------------------------------------------ | | **Caller** | DSUL automations | AI agents (LLM) | | **Tool discovery** | Manual (you specify which automation) | Automatic (agent selects tools) | | **Authentication** | Workspace config | Headers, user session, or workspace config | | **Use case** | Deterministic workflows | Conversational AI, agentic tasks | ### Provider-Specific Fields Some external services expose configurable fields that are defined outside Prisme.ai, such as Jira custom fields, CRM properties, ticket attributes, or board columns. Connector operations should expose these through a generic structured object such as `fields` when the provider API accepts arbitrary field IDs or keys. Prisme.ai passes values to fields that already exist in the external service; it does not create new provider fields unless the connector explicitly includes administrative tools for that provider. When field IDs, available work types, screens, layouts, or create/update metadata matter, use the connector's discovery operations first. Avoid hardcoding provider-specific IDs or assumptions in automations and tests. ### When to Use App Mode * You're building a workflow automation (e.g., sync data on a schedule) * You know exactly which operations to call * You want to chain multiple operations in a single automation * You don't need AI to decide which operations to use ```yaml theme={null} # Example: App mode in DSUL - Jira.searchJira: jql: "project = DEV AND status = 'In Progress'" maxResults: 10 output: results ``` ### When to Use MCP Mode * You're building an AI agent that needs access to external services * The agent should dynamically choose which operations to call * Users interact via natural language (chat interface) * You want tool discovery and schema introspection ```json theme={null} // Example: MCP tool call from AI agent { "name": "searchJira", "arguments": { "jql": "project = DEV AND status = 'In Progress'", "maxResults": 10 } } ``` ## Architecture Each connector is built as a **single Prisme.ai workspace** and published as an app. App-mode instructions and workspace-specific MCP calls run in the installed tenant app instance, reading that workspace's configuration and credentials. Connectors using central per-user OAuth can also expose an optional global MCP endpoint from the connector core. ```mermaid theme={null} flowchart TB subgraph tenant["Your Workspace"] direction TB dsul["DSUL Automation"] agent["Agent Factory agent"] subgraph inst["Connector app instance (tenant context)"] mcp["MCP endpoint"] app["App-mode instructions"] end end core["Optional global MCP endpoint
(connector core)"] api["External API
(Jira, Microsoft Graph)"] dsul -->|"Connector.operation()"| app agent -->|"MCP tools"| mcp agent -.->|"Optional global MCP"| core mcp --> api app --> api core -.-> api ``` The connector wraps the external API as reusable operations, handles authentication (API key, OAuth2, JWT, and more), and groups endpoints into LLM-friendly entity tools. Workspace-specific MCP endpoints gate agents with the tenant allowlist; optional global endpoints rely on per-user OAuth and provider authorization. Review execution topologies, request controls, authentication, secrets, and agent authorization. ## Next Steps Get started with Jira integration Create your own connector for any API