MCP connections for Claude Code and Claude Desktop

1. An owner switches the connector on under Connections in Selan. Every connector starts off. A few need something from you first, and the page asks for it: an OAuth client you create for Google, the address of your Outline, the bot and chat Telegram posts with, or Slack's Add to Slack (or your own Slack app's token).
2. Each developer picks where it runs on their own Connections page: Claude Desktop, the Claude CLI or both. Desktop starts on and the CLI starts off. The connector appears the next time they start that app. For the CLI, Selan adds it to ~/.claude.json and removes it again if the company switches it off.
3. The first time they use it, developers sign in as themselves, so their own permissions apply and Selan never holds their account. Owners can see what is offered, but not who has signed in.

Most connectors are the provider's own MCP server on the provider's own host. Selan runs the ones on mcp.selan.ai: its own two, and providers that have no hosted server or none Claude can sign in to directly. Reactor remote agents use the same catalog and sign in the same way (see Selan MCP).

The Sign-in column says how a developer authenticates. OAuth means the provider's own sign-in page. Selan sign-in means the developer's Selan session is enough.

Code and engineering

Google

Projects, docs and design

Messages and meetings

Product analytics and marketing

Payments, finance and people

Selan and the web

Missing a tool your team uses? Write to info@selan.ai.

Slack lets Reactor agents read Slack and post in it as your company's own bot, never as a person. It works in Slack Connect channels too.

Add to Slack (recommended)

In Connections → Slack, click Add to Slack — install the Selan app and approve. Once installed, the button reads Reconnect Slack. Then /invite @Selan Agents into each channel agents should use.

Or use your own Slack app

Under the button, open Use your own Slack app instead.

1. Open api.slack.com/apps → Create New App → From a manifest, pick your workspace, paste the manifest below, then Create.

   display_information:
     name: Selan Agents
     description: Selan's remote agents read and post here as your company's bot.
   features:
     app_home:
       home_tab_enabled: false
       messages_tab_enabled: true
       messages_tab_read_only_enabled: true
     bot_user:
       display_name: Selan Agents
       always_online: true
   oauth_config:
     scopes:
       bot: [chat:write, chat:write.customize, reactions:write, channels:history, groups:history, im:history, mpim:history, im:write, im:read, mpim:read, channels:read, groups:read, users:read, users:read.email]
   settings:
     org_deploy_enabled: false
     socket_mode_enabled: false
     token_rotation_enabled: false

2. Click Install to Workspace → Allow.
3. Copy the Bot User OAuth Token (OAuth & Permissions, it starts xoxb-) and the Signing Secret (Basic Information → App Credentials), paste both under Connections → Slack, and save. Selan checks them with Slack and keeps them encrypted; no developer or agent ever holds them.
4. In Slack, /invite @Selan Agents into each channel agents should use.

Using an existing app instead? It needs the scopes above, chat:write.customize for agent names, and the Messages Tab on for DMs. Then reinstall it.

What a post looks like

The sender is the agent's name, like Kepler Deploy, shown with the chat:write.customize scope. The body is Markdown. Under it is one footer line: Posted by Kepler Deploy for @person · run, linking the run that sent it. A person's own post shows Posted by @person.

Tools

• slack_post_message: post in a channel, or reply in a thread. Pass a user ID (U…) as the channel to DM that person.
• slack_update_message: replace the text of the agent's own post.
• slack_delete_message: delete the agent's own post.
• slack_add_reaction: add an emoji reaction.
• slack_remove_reaction: remove the bot's reaction.
• slack_read_channel: a channel's recent messages, oldest first, with names and times.
• slack_read_thread: a thread, its first message then the replies.
• slack_read_dm: the bot's direct messages with the person the run acts for.
• slack_find_user: find a person by exact email, or by a name fragment, and get their user ID.
• slack_user_info: one person by user ID: name, title, timezone, whether a bot.
• slack_find_channel: find a channel the bot can see by a name, topic or purpose fragment, or by exact #name.
• slack_channel_info: one channel by ID: name, privacy, whether the bot is in it, member count, topic and purpose.

Limits

• The bot works only in channels it has been invited to.
• Sending DMs needs the app's Messages Tab on, or Slack refuses with messages_tab_disabled.
• There is no message search: Slack does not allow it with a bot token.
• An agent can edit or delete only its own posts.

Before a Reactor agent can sign in to Jira & Confluence, an admin of your Atlassian organization must allow Reactor once. Atlassian's MCP server signs in only to addresses the organization allows, and Reactor's is not on its default list. Claude Code and Claude Desktop are, so developers need nothing.

Until then, Atlassian's sign-in page says Access to this domain is restricted and Accept stays greyed out.

Allow Reactor

1. An admin of your Atlassian organization opens admin.atlassian.com and picks the organization.
2. Go to Rovo → Rovo MCP server and, under Your domains, click Add domain.
3. Enter https://reactor.selan.ai/api/connectors/callback and click Add.

Then sign the agent in again: Connect on its Connectors page in Reactor, or open the link that selan_remote_agent_connector_connect returns. Atlassian's own steps are in Control Atlassian MCP server settings.

• Browser Extension: the connector that reaches your own Chrome.
• Selan MCP: your company's Selan data and remote agents over MCP.