Skip to main content

🤖 AI Agent Connect: Vlge MCP Server

[BETA] You can connect your agent to your Vlge instance

⚡️ Note: Vlge Pro required | Advanced Use only | Beta Version

You can connect either Claude Desktop or ChatGPT (or other if compatible) to your Vlge instance as an admin to gather data or use several tools to make changes or update your community. Our MCP Server follows your account's permissions with regards to what you can read or edit in Vlge. Not all features or functions are supported and the feature is in limited beta.

See our Developer Site additional instructions and an up to date list of all tools available.

⚠️ Anyone with a Vlge portal account can connect to the MCP server and use Agents to connect to available information in the portal. However, at this time, the ONLY supported method for using Vlge is through the portal itself when logged in. The help center does not have additional documentation on using the MCP server.


Safety and limits

  • Every write is recorded in the community audit log as you, tagged with the assistant that made it.

  • Deleted meetings, polls, forms, and help sections go to Rewind and can be restored by an admin for 45 days.

  • Community emails are draft-only from an assistant: vlge_create_email_draft stages the message in Messaging Center and you send it from Vlge.

  • Anything that reaches residents (for example vlge_send_meeting_notice) is a two-step handshake: the first call returns CONFIRMATION_REQUIRED with recipient counts and a single-use token; the assistant must show you that summary and call again with confirmation_token only after you approve. Changing any argument invalidates the token.

  • Each connection is limited to 60 calls per minute, 3,000 per day, and 20 writes per minute. Exceeding a limit returns RATE_LIMITED with a retry time.

  • Access tokens last one hour; refresh tokens rotate on every use and expire after 30 days of inactivity.

  • Disconnect an assistant any time from Profile → Security → Connected AI apps in Vlge. Access ends immediately.

Seeing what an assistant did

  • Audit log. Community admins open Admin Tools → Audit Log and set Client to AI assistants. Every MCP write appears there under your name with the connector identified (for example MCP · Claude), alongside any action an assistant took through the regular API with your browser session.

  • AI connections. Admin Tools → AI connections lists every assistant connected to the community, who connected it, the scopes it holds, and when it was last used. Admins can revoke a connection there.

  • Your own connections. Profile → Security → Connected AI apps shows the assistants you have authorized and lets you disconnect them.

  • Assistants using the REST API directly. A desktop assistant that drives Vlge with your signed-in browser session instead of this connector is still identified by its user agent and tagged in the same audit log. Those calls are limited to 60 writes per 10 seconds and 300 per minute per user and get the same CONFIRMATION_REQUIRED handshake on resident-facing actions. The MCP connector is the supported path.


🔑 Two ways to connect an AI agent to Vlge

Vlge's AI connection (the MCP server at https://mcp.vlge.io/mcp) accepts two kinds of connection. Which one you use depends on whether a person will be there to sign in.

Option 1: Sign in with your Vlge account (OAuth). For people using Claude, ChatGPT, or another AI assistant

Use this when you're chatting with an assistant yourself.

  • Add Vlge as a connector in your assistant using the server URL above.

  • Sign in with your normal Vlge account, pick one community, and approve the permissions the assistant should have.

  • The assistant acts as you. It can never do more than your role in that community allows.

  • You don't need an API key, and there's nothing to copy or rotate.

  • Every change it makes appears in the community audit log under your name, with the assistant identified (for example, "MCP · Claude").

  • You can disconnect it anytime under Profile → Security → Connected AI apps. Community admins can see and revoke connections under Admin Tools → AI connections.

Option 2: Service key. For scheduled bots and agents that run on their own

⚠️ Caution: For advanced users only. Using automated and un-monitored bots with write permissions for your instance is a risk. You should only use this method if you are familiar with the capabilities and approaches to protect against unwanted changes. Many changes that can be made by the automated agent are not recoverable.

Use this for a bot that runs on a schedule with nobody there to sign in, such as an ARC screening bot or a ticket triage bot.

  • A community member with the Configure integrations permission creates a service key at Developer → Keys on developer.vlge.io. The key must be for one community only.

  • The person creating the key picks the key's scopes (for example, read ARC requests or comment on tickets) and its audience: Member, Board, or Admin.

  • Member sees what every resident sees. Board adds board-only meetings and documents, and requires board document access. Admin sees everything in the community, including admin-only folders and the election administrator view (turnout and voter roll), and requires full admin access in that community. No key, at any audience, can reach another community or the Vlge platform.

  • The bot sends the key as a request header: Authorization: Bearer vlge_svc_…

  • The key acts on behalf of the person who created it, and that person's permissions are checked on every call. If that person loses access to the community, the key stops working.

  • Service keys can't use the personal "My account" tools, community content tools (FAQs, groups, polls, forms, binder), or anything that sends messages to residents. Writes made with a service key never email anyone.

  • Changes appear in the audit log under the key's name (for example, "MCP · service:arc-screening-bot").

  • Revoking a key at Developer → Keys stops it within about 15 seconds.

  • Personal API keys (starting vlge_live_) and keys that cover more than one community can't be used with the AI connection.

  • The Organization API must be turned on for the community. To make changes (not just read), Organization API writes must also be on; otherwise write tools return WRITES_DISABLED.

Safety and limits

  • Assistants can't move money, pay bills, post charges, or issue new violation notices.

  • Anything that reaches residents, such as sending a meeting notice, requires the user to confirm first. Community emails from an assistant are saved as drafts only.

  • Deleted meetings, polls, forms, and help sections go to Rewind and can be restored for 45 days.

  • Signed-in connections are limited to 60 calls per minute, 3,000 per day, and 20 writes per minute. Service keys use their own API key limits.

Which one should I use?

  • You're chatting with Claude or ChatGPT yourself: sign in with your Vlge account (Option 1).

  • A bot runs on a schedule without you: use a single-community service key (Option 2).

Full setup steps, the list of permissions, every available tool, and error codes are at

Scheduled bot setup is in the section

🛟 Need more help? Just ask!

Did this answer your question?