Connector Catalog

Connect AI agents to external tools and work with data, content, and workflows directly from your conversation.

Key benefits and use cases

The Connector Catalog lets you connect AI agents to external tools. Depending on the connector, an agent can search and analyze data, read and edit content, or take action.

The available actions depend on your connected account, your permissions, and the tools enabled by the provider.

There are currently 29 connector presets available:

• Data and content: Airtable, Box, Contentful, Dropbox Dash, Google Workspace, Notion, WordPress.com

• Development and infrastructure: Cloudflare, GitHub, LaunchDarkly, Sentry, Supabase

• Sales, marketing, and CRM: HubSpot, Kickscale, Moco, PostHog, Typeform

• Communication and collaboration: Atlassian Rovo, Calendly, Intercom, Microsoft 365, Miro, Slack, Zoom

• Analytics and other tools: Apify, Bitly, Metabase, Mobbin

Figma, Pitch, and Canva aren't available in awork yet. Work on these integrations is underway.

How to use the Connector Catalog

A connector can share data with an external tool. Where supported, an agent can also create or modify data there. The available actions correspond to the permissions you grant to your connected account.

Connect a connector

1. Open the Library

   • Go to Einstellungen (the gear icon at the bottom left of the main menu) > Library
   • Or open AI in the main menu on the left and click the Library button in the top right

2. Select Konnektoren in the left navigation to view the available connections.

3. Then click the + Konnektor button at the top to add a new connector.

4. Select a connector from the catalog
   The Konnektor einrichten modal opens with a catalog of available integrations. You can use the search field to find a specific connector, or scroll through the tiles. Select the connector you want by clicking its tile.

5. Fill in the connector settings shown.

6. Add a connection.

7. Sign in with the relevant provider and confirm access.

You've now selected a connector from the catalog and can continue setting up the connection.

Most connectors only require you to sign in. You can find additional settings in the details for each connector.

Permissions and access

The actions a connector can perform depend on more than just the connector itself. Your connected account's permissions, the access you grant during authorization, and any settings configured by your organization or administrator also matter.

Preconfigured connectors

Airtable

With Airtable, an agent can search and analyze your base data, create and update records, and work with bases, interfaces, and automation drafts. Your existing Airtable permissions remain in place. Users with read-only access can't make changes.

Permissions: The connector can read and modify records, comments, and the base structure in the bases you authorize.

Requirements: An Airtable account, data in bases, and permission to use third-party integrations in your Airtable organization. An organization admin may need to allowlist the integration.

Airtable documentation

Apify

With Apify, an agent can find and run Actors. These are ready-made tools in the Apify Store that collect and process web data. The agent can also retrieve results and search the Apify documentation.

Permissions: The connector gets access to your Apify account to run Actors and read or manage their data and results within the permissions of your account.

Requirements: An Apify account and authorization for the hosted connection.

Apify documentation

Atlassian Rovo

Atlassian Rovo brings information from Jira, Confluence, Compass, and Teamwork Graph into your conversation with an agent. You can search for work items, summarize pages, and create content using your existing Atlassian permissions.

Permissions: The connector can read and modify Jira work items, Confluence content, and supported Compass and Teamwork Graph data within the permissions of your account.

Requirements: An Atlassian Cloud account with access to the relevant Jira, Confluence, or Compass apps. For OAuth, your organization must allow the AI tool's domain. Organization-wide IP allowlists and authentication policies can also block requests. API token authentication is an advanced option that depends on an organization setting.

Atlassian Rovo documentation

Bitly

With Bitly, you can manage link workflows through natural language requests to the agent. Access is based on your Bitly account.

Permissions: Access is determined during authorization and may include managing your links. The permissions shown during authorization determine what the connector can do.

Requirements: A Bitly account and authorization for the hosted connection.

Bitly documentation

Box

With Box, an agent can search for and work with files and folders. The connector also provides access to Box AI tools. Access remains limited by your existing Box permissions.

Permissions: Access depends on the permissions configured for the Box integration and your access to files and folders. The permissions shown during authorization determine what the connector can read or modify.

Requirements: A Box admin must enable MCP in the Box Admin Console. For custom clients, Box Integration Credentials are also required, including an OAuth Client ID, Client Secret, Redirect URI, and appropriate scopes such as Content Actions. All requests remain subject to Box permissions.

Box documentation

Calendly

With Calendly, an agent can find availability, book or cancel meetings, and create scheduling links. It can also list your event types. Actions depend on the permissions of your Calendly account.

Permissions: The connector can read scheduling information and perform actions such as booking and canceling meetings.

Requirements: An active Calendly account. Authorization takes place through the hosted connection. A pre-registered Client Secret or personal Access Token isn't supported.

Calendly documentation

Canva

Canva isn't available in awork yet. Canva currently restricts its MCP server to certain clients. Integration with awork is in progress.

Canva documentation

Cloudflare

With Cloudflare, an agent can work with DNS, Workers, R2, and Zero Trust, among other services. It can read account settings, suggest changes, and apply changes within the permissions you've granted.

Permissions: The Cloudflare account permissions selected during authorization determine which services the connector can read or modify.

Requirements: Sign in to Cloudflare using OAuth with the permissions selected during authorization. For CI/CD or automation, you can provide a Cloudflare API token with the required permissions as a bearer token. User and account tokens are supported.

Cloudflare documentation

Contentful

With Contentful, an agent can read and edit content, manage content models and assets, and work with spaces and environments. The available actions depend on your Contentful permissions and the tools enabled for the space and environment.

Permissions: The connector follows your Contentful permissions and the tools your administrator has enabled for each space and environment.

Requirements: A Contentful account with access to the spaces and environments you want to use. The Contentful MCP app must also be installed and configured for remote use in each environment. To meet data residency requirements, choose the appropriate global or EU endpoint. A local server instead requires a Content Management API Personal Access Token. Space and environment defaults are optional.

Contentful documentation

Dropbox Dash

With Dropbox Dash, an agent can search your connected business sources and read documents and files. The available search features are read-only.

Permissions: The connector requests access to your identity and permission to read and write Dash content. The documented tools currently focus on finding and reading information.

Requirements: A Dropbox Dash account with connected sources. Dropbox requires authorization for the hosted connection.

Dropbox Dash documentation

Figma

Figma isn't available in awork yet. Figma currently restricts its MCP server to certain clients. Integration with awork is in progress.

Figma documentation

GitHub

With GitHub, an agent can work with repositories, issues, pull requests, and discussions. Depending on your account and the available tools, it also supports development workflows, automated checks, and code security.

Permissions: The connector requests permissions to read and modify repositories and projects, download or publish packages, work with gists, and manage notifications. Actual access is limited by your account permissions and the permissions of the installed GitHub app.

Requirements: A GitHub account and authorization for the hosted endpoint. The upstream documentation doesn't list any additional client ID, tenant, or token fields in the catalog. Access and available actions depend on the GitHub account you're signed in to and the selected toolset or read-only URL.

GitHub documentation

Google Workspace

With Google Workspace, an agent can work with Calendar, Docs, Sheets, and Slides. Depending on the tools enabled, it can manage events and read and edit documents, spreadsheets, and presentations.

Permissions: The connector requests permission to manage calendars and read or modify documents, spreadsheets, and presentations. It also requests basic profile and email address information.

Google Workspace (advanced)

With Google Workspace (advanced), an agent can work with Gmail, Calendar, Drive, Docs, Sheets, and Slides. Depending on the tools enabled, it can manage events, retrieve and send emails, and read and edit files, documents, spreadsheets, and presentations.

Permissions: The connector requests permission to manage calendars and read or modify emails, files, documents, spreadsheets, and presentations. It also requests basic profile and email address information.

Setup: To use the advanced Google Workspace connector, you'll need your own OAuth client in Google Cloud.

1. Configure OAuth consent: Under Google Auth Platform → Branding, enter an app name and a support email address. Under audience, select Internal.

2. Add scopes: Under Data Access, add the scopes required by the connector:

openid
https://www.googleapis.com/auth/userinfo.email
https://www.googleapis.com/auth/userinfo.profile
https://www.googleapis.com/auth/gmail.modify
https://www.googleapis.com/auth/drive
https://www.googleapis.com/auth/calendar
https://www.googleapis.com/auth/documents
https://www.googleapis.com/auth/spreadsheets
https://www.googleapis.com/auth/presentations

3. Create the OAuth client: Under Clients → Create client, select Web application, enter a name, and add the awork redirect URI:

https://app.awork.com/api/v1/agents/connectors/auth/callback

Create the client and store the Client ID and Client Secret securely. Use the ID and secret in the advanced Google Workspace connectors in awork.

HubSpot

With HubSpot, an agent can query CRM records, sales activities, marketing content, campaigns, conversations, and account information. It can create or update selected CRM records and activities, publish landing pages and blog posts, manage campaigns, and prepare marketing email drafts. The available actions depend on your HubSpot permissions.

Permissions: The connector's permissions depend on your HubSpot user permissions, your account subscription, and the access you grant during authorization.

Requirements: An MCP Auth app must be set up in your HubSpot account. Authorization takes place through the hosted connection. The available scopes depend on the signed-in user's permissions and the account's subscriptions.

HubSpot documentation

Intercom

With Intercom, an agent can search and retrieve workspace data such as conversations and contacts. The documented tools also support Intercom actions such as creating Help Center articles. Access is based on the authenticated user.

Permissions: The connector requires access to contacts, companies, and conversations. The documented permissions also allow it to create internal conversation notes and create or update Help Center articles.

Prerequisites: According to Intercom, the hosted MCP server currently supports only US-hosted workspaces. Each connection is authenticated. No separate provider credentials or additional setup steps were visible on the available first-party page.

Intercom documentation

Kickscale

With Kickscale, an agent connects to Sales Insights, allowing you to bring conversation and deal context into your workflows. In its extended integrations, Kickscale combines CRM, meeting, email, and calendar data to analyze customer conversations, coaching signals, and deal progress.

Permissions: Access is determined when you authorize the Kickscale connection. You can see which information and actions are available during authorization.

Prerequisites: The linked page confirms that an MCP server is available, but does not document authentication, account requirements, or setup steps.

Kickscale documentation

LaunchDarkly

With LaunchDarkly, an agent can manage feature flags and targeting rules using natural language. It can also manage AgentControl configurations and query observability data such as logs, traces, errors, metrics, and dashboards.

Permissions: The connector requests read and write access, as well as access to observability data. The actions available depend on your LaunchDarkly permissions.

Important: You must enable the hosted MCP server in LaunchDarkly before using it. The hosted server is not available in the Federal and European Union environments. You must use the local server there.

LaunchDarkly documentation

Metabase

With Metabase, an agent can access governed business data, create and run queries, and return results as text or as inline bar charts, line charts, or tables. It can also provide database metadata, so an agent can prepare Metabase questions and dashboards using real tables and fields.

Permissions: Depending on your Metabase configuration, the connector can access available business data, run queries, and create or modify saved questions, metrics, dashboards, and collections. Your Metabase permissions and administrator settings determine which actions are actually available.

Prerequisites: A Metabase administrator must first enable AI features, then enable the MCP server under Admin > AI > MCP. Each connection uses the permissions of the signed-in Metabase user.

Required configuration: Subdomain: The part of your account address before .metabaseapp.com. For example.metabaseapp.com, enter example. The value can contain letters, numbers, and hyphens and must be no more than 63 characters. It must start and end with a letter or number. Enter only the subdomain, without https://, dots, or paths. Then sign in and confirm access.

Metabase documentation

Microsoft 365

With Microsoft 365, an agent can work with Outlook email and calendars, OneDrive files, Excel workbooks, OneNote, To Do, Planner, contacts, and Microsoft Search. Depending on the enabled tools and account access, it can also work with Teams, SharePoint, meeting transcripts and recordings, and shared mailboxes and calendars.

Permissions: The connector requests access to read and modify email, calendars, contacts, and files, as well as to send messages and read SharePoint content. Depending on your account permissions, Teams permissions may also include creating or deleting channels and changing team membership.

Prerequisites: The connected upstream server requires Microsoft authentication and requests the appropriate Microsoft Graph permissions based on the enabled tools. The default app supports interactive authentication. For a custom production deployment, you can instead use an Azure Entra app registration with a Tenant ID, Client ID, optional Client Secret for public apps, and the appropriate redirect URIs.

Microsoft 365 documentation

Miro

With Miro, an agent can search and summarize boards, build boards, create diagrams, and work with comments. Supported content includes frames, sticky notes, shapes, text, cards, documents, tables, and images from public URLs.

Permissions: The connector can read and modify boards and their contents that you have access to. It can also access your identity and email address.

Prerequisites: During authorization, you select a Miro team. The connector can access only boards shared with the user under their current permissions. The server is available on all Miro plans. In an Enterprise environment, an administrator may need to enable it for the organization or selected teams.

Miro documentation

Moco

With Moco, an agent can query business data about leads, offers, projects, planning, time tracking, billing, customers, and business development. If write access is enabled, it can also create or update projects, contacts, planning entries, time entries, leads, and offers, among other things.

Permissions: The connector can read the data you have access to in Moco. If your administrator enables write access, it can also create or modify supported records.

Prerequisites: An account administrator must enable the AI feature under Settings > Account > AI and select read-only or read/write access for MCP. Users then authorize through an OAuth-enabled client using their personal Moco login. Existing Moco permissions remain in effect. Depending on the client, a paid plan or admin rights may be required.

Required configuration: Subdomain: The part of your account address before .mocoapp.com. For example.mocoapp.com, enter example. The value can contain letters, numbers, and hyphens and must be no more than 63 characters. It must start and end with a letter or number. Enter only the subdomain, without https://, dots, or paths. Then sign in and confirm access.

Moco documentation

Mobbin

With Mobbin, an agent can access a searchable library of real product screens and design patterns. It can find examples of flows such as onboarding, checkout, pricing, and permissions, compare solutions from published products, and use these references when designing products.

Permissions: The connector requests access to your login, profile, and email address, as well as permission to keep the connection active. Access to the design library is governed by your Mobbin account.

Requirements: Mobbin paid plan. Sign in with your Mobbin account and confirm the connection in your browser. According to Mobbin, no API key is required.

Mobbin documentation

Notion

With Notion, an agent can read and update pages and other content that you can access in the selected workspace. It can find workspace knowledge, summarize or organize content, create and update pages, add comments, and upload files within Notion's limits.

Permissions: The connector uses the workspace access you grant. This may include permission to read and edit content. Your Notion permissions determine which pages and other content are accessible.

Requirements: A Notion account with access to the desired workspace. The connection uses a browser-based OAuth flow. During the process, you choose which workspace to connect. Notion currently requires interactive authorization.

Notion documentation

Pitch

Pitch is not yet available in awork. Pitch currently restricts its MCP server to certain clients. Integration with awork is in progress.

Pitch documentation

PostHog

With PostHog, an agent can work with product analytics and development data. This includes analytics and HogQL queries, errors and stack traces, feature flags and experiments, support tickets, and CDP destinations.

Permissions: The connector requests broad access to analytics, feature flags, experiments, recordings, data sources, and project and organization settings. Many of these permissions allow the agent not only to view data and settings, but also to change them.

Requirements: Sign in to the PostHog organization that contains the desired project and confirm OAuth access. Some tools that use PostHog AI require AI data processing to be enabled and may incur PostHog AI usage costs.

PostHog documentation

Sentry

With Sentry, an agent can search for errors, analyze performance, triage issues, read Sentry documentation, and manage projects. This lets you investigate errors and performance issues using up-to-date Sentry context.

Permissions: The connector requests access to organization information, as well as permissions to create or modify projects and teams and update issues.

Requirements: A Sentry account with access to the required organization or project. The first time you connect, a browser-based OAuth flow starts.

Sentry documentation

Slack

With Slack, an agent can search messages, files, people, and channels, and read channel or thread history. It can also send or draft messages, create conversations, add reactions, manage Canvases and Lists, and upload files. The available actions depend on the granted scopes and workspace permissions.

Permissions: The connector can search and read accessible channels, direct messages, files, and profiles. It can also send messages, manage conversations, edit Canvases and Lists, and upload files.

Requirements: Register a Slack app or use an existing one, and set up confidential OAuth with a client ID and client secret. Slack currently only allows apps published in the Marketplace and internal apps to use the Slack MCP server. Unpublished apps are not supported. Add the scopes for the actions you want to use, and get approval from your workspace administrator if needed.

Required configuration: First register an OAuth app with the provider, then fill in these fields.

Slack documentation

Supabase

With Supabase, an agent can inspect and work with Supabase projects and Postgres databases. Depending on the enabled tool groups, it can query or migrate databases, check logs and advisors, generate TypeScript types, deploy Edge Functions, manage projects and branches, and search Supabase documentation.

Permissions: The connector requests permissions to read and modify projects, databases, Edge Functions, environments, and storage. Depending on your account permissions, it may also access analytics and project API keys or secrets.

Requirements: A Supabase account and selection of the organization that contains the desired project. The hosted server normally uses browser-based OAuth with dynamic client registration. For production data, Supabase recommends restricting access to the desired project and using read-only mode.

Supabase documentation

Typeform

With Typeform, an agent can create and edit forms, including their fields, logic, and publishing settings. It can also manage automations, contacts, and lists, and analyze responses using raw data, aggregates, rankings, and time series.

Permissions: The connector can read and modify forms, responses, contacts, automations, webhooks, and workspaces. It can also access account information and insights.

Requirements: A Typeform account. The first time you connect, an OAuth 2.0 flow starts. EU-hosted Typeform accounts use different endpoints. Typeform also requires the redirect domain of a hosted connector to be allowlisted before you can connect.

Required configuration: Data region: Choose the region used by your Typeform account: Global (api.typeform.com, default), EU (api.eu.typeform.com), or EU2 (api.typeform.eu). Choose the region for your existing account. This setting does not move your data. Then sign in and confirm access.

Typeform documentation

Webflow

With Webflow, an agent can work directly with your Webflow projects. It can create and edit responsive pages, elements, styles, components, variables, assets, CMS Collections, and custom code. It can also read site analytics, while respecting your existing Webflow role and permissions.

Permissions: The connector uses the site or workspace access you grant. What it can read or change depends on your Webflow role and the permissions granted during authorization.

Requirements: A Webflow account with access to at least one site. Sites are authorized through a browser-based OAuth flow. During this process, Webflow installs the MCP Bridge App.

Webflow documentation

WordPress.com

With WordPress.com, an agent can edit your connected websites using natural language. It can use tools for website content and administration. For eligible self-hosted sites, the same connection can also provide Jetpack features such as Backup, Monitor, Newsletter, and Scan.

Permissions: The connector requests extensive access to your account and websites, including content, users, comments, media, and statistics. The actions available depend on your WordPress permissions and the connector tools you've enabled.

Requirements: MCP must be enabled in your WordPress.com account. MCP is available on paid WordPress.com plans, during the first 30 days of a newly created free site, and on self-hosted sites connected through Jetpack with a Jetpack AI or Jetpack Complete plan.

WordPress.com documentation

Zoom

With Zoom, an agent can find meeting information and work with Cloud Recordings, Zoom Docs, and Zoom Hub. Usage depends on your existing permissions and Zoom licenses.

Permissions: The connector can find and read meeting content, recordings, documents, and notes. It also requests permissions to create, edit, or delete meetings, change meeting settings, manage document collaborators, and write content to Zoom Hub.

Requirements: A Zoom account with admin or developer authorization, as well as the required licenses for the Zoom products you want to use. Create a General App in the Zoom App Marketplace, add the required OAuth scopes, and obtain a Client ID and Client Secret. Zoom's remote MCP servers require manual client registration and OAuth with PKCE. Dynamic client registration isn't supported.

The connector uses Zoom's current Unified Streamable Endpoint: https://mcp.zoom.us/mcp/zoom/streamable.

Required configuration: First, register an OAuth app with the provider, then fill in these fields.

Zoom documentation

Notes

• A connector doesn't automatically expand your access: An agent works within the permissions allowed by your connected account and the provider. This also applies to actions that can change data.

• Enter specific settings correctly: For Metabase and Moco, use the subdomain from your account address. For Typeform, select the Data region for your existing account. Slack and Zoom also require an OAuth Client ID and Client Secret.

FAQ

Is adding a connector to the library enough?

No. After adding it, you need to set up a connection, sign in with the relevant provider, and approve access.

Can an AI agent change data using connected tools?

When a connector supports write actions, an agent can create or change data. The available actions depend on your account permissions, the access you've granted, and the provider tools you've enabled.

Can an AI agent access all my Google Drive files?

No. The Google Workspace connector can currently retrieve only Google Drive files created by an agent. Other Google Drive files aren't accessible.

Do the available actions depend on my account?

Yes. The available actions depend on the connected account, its permissions, and the tools enabled by the provider.

What should I do if the connection isn't working?

First, check whether the connector is available in awork, whether the required provider settings are enabled, and whether your account has the necessary permissions. For connectors with additional fields, also check the configuration values you entered.

What permissions do connectors have?

The permissions required vary by connector. They determine what data the connector can read or change. Your account permissions and your organization's or administrator's settings may also restrict access.

Can a connector change data too?

Yes. Many connectors support changes in addition to read access. The actions actually available depend on the connector, your account permissions, and the access granted during authorization.