Agents
Connect an AI assistant to your Nuvelics organization over MCP, and see what it can do once it is connected.
Nuvelics runs an MCP server, the Model Context Protocol endpoint an AI assistant connects to. Once an assistant is connected, you can ask it in plain language to review your organization's pitches, read stories, ask for outputs such as an article or a social post, and push a story to a newsroom content system, and it does the work through Nuvelics's own tools instead of through the web app or the command line.
A connection you approve acts as you, in one organization you choose, limited to what you are allowed to do there. Nothing it does reaches beyond that. A script or unattended job can use a static API key instead, which acts for the organization; see Without a browser.
The path through this section
- Before you connect: an owner or admin turns MCP on for the organization, and you learn what a connection acts as.
- Connect a client: the server address, adding it in your assistant, signing in and approving the connection.
- What a connection can do: the limits on a connection, how they change with your role, and how to disconnect.
- Without a browser: a static API key for a script or an unattended job, and the tools it cannot use.
- Slackbot: connecting Slack's built-in assistant.
- Troubleshooting: what to do when a call fails or a connection stops working.
- Reading these docs as an agent: the machine-readable forms of this site, and its own MCP server.
When you know what you are connecting and want to see exactly what it can call, the tool reference lists every tool, its parameters and its behavior, grouped by what it works on.