Skip to content
Request early access

Integrations

Connect the tools you already use

A connector is a tool an agent can reach, nothing more. Commt scopes every connector to the agent it's granted to, so adding Slack to one agent doesn't hand every agent in your account a door into your workspace. Twenty connectors are available at launch, covering messaging tools, work tools and customer systems, plus Zapier for anything beyond that list and a direct API key for your own product.

What a connector does, and what it doesn't

A connector is not a standing link between Commt and your account at some other tool. It's a capability granted to one specific agent: which tool, which resource inside that tool (a channel, a project, a folder), and whether the agent can read it, write to it, or both. An agent with no connectors granted can't reach any outside tool at all, the same way an agent with no knowledge base bound to it has nothing to retrieve. Reach is opt-in, connector by connector, agent by agent, not a menu of everything the platform happens to support.

Can I limit which tools an agent can reach?

Yes, that's how connectors are built to work on Commt. The question worth asking about any AI agent platform isn't how many tools it connects to, it's whether an agent reaches only what you scope to it or everything the account has access to. Commt's answer is the first: granting Slack to a support agent doesn't grant it to a sales agent in the same account, and granting read access to a Drive folder doesn't grant write access unless that's set separately. The twenty connectors below are the reach available to grant, not twenty doors standing open by default.

Twenty connectors available at launch

Grouped by what they connect an agent to, not by direction. Each line below is what an agent scoped to that connector is built to do with it.

Messaging and email

  • Slack: reading and posting in the channels you scope to an agent
  • Microsoft Teams: reading and posting in the channels and chats you scope to an agent
  • Telegram: replying in a bot conversation or channel scoped to an agent
  • WhatsApp: replying to messages on a WhatsApp Business number scoped to an agent
  • Instagram: replying to comments and direct messages on an account scoped to an agent
  • Facebook: replying to comments and messages on a Page scoped to an agent
  • Reddit: replying to posts and comments on a subreddit or account scoped to an agent
  • Gmail: reading and sending from an inbox scoped to an agent
  • Outlook: reading and sending from an inbox scoped to an agent

Work and knowledge tools

  • Jira: reading and updating issues in the projects you scope to an agent
  • Confluence: reading and writing pages in the spaces you scope to an agent
  • Google Docs: reading and writing documents you scope to an agent
  • Google Drive: reading files in the folders you scope to an agent
  • Notion: reading and writing pages in the workspace you scope to an agent
  • Airtable: reading and writing records in the bases you scope to an agent
  • GitHub: reading issues, pull requests and code in the repositories you scope to an agent

Customer and sales tools

  • Salesforce: reading and updating records in the objects you scope to an agent
  • HubSpot: reading and updating contacts and deals you scope to an agent
  • Zendesk: reading and responding to tickets in the queues you scope to an agent

Or connect another way

  • Zapier: reaching tools beyond this list, scoped the same way as any connector above
  • API key: calling the finished agent directly from your own product

Ground an agent in the tools it needs, and nothing else

Which connectors an agent can reach is set when it is built, alongside the knowledge bases it's allowed to retrieve from. Neither is required: an agent can run on a model alone, with no connector and no knowledge base bound to it, if that's all the job needs.

Request early access