Using the Bolti Voice AI MCP Server in Cursor (2026 Guide)

Bolti is a voice AI platform for building conversational phone agents that helps you deploy production-ready voice workflows with sub-second latency. For developers, Bolti is built to be managed directly from your code editor using our official Model Context Protocol (MCP) server. You can sign up for a create your developer account to get 50 free minutes and start prototyping immediately.

By exposing our entire platform—including agents, tools, phone numbers, and call logs—as a set of tools for LLMs, Bolti makes voice agent development a first-class developer experience. This guide will show you how to connect the Bolti voice AI MCP server to Cursor and build voice workflows without leaving your editor.

What is the Bolti voice ai mcp server?

The Bolti voice AI MCP server is an implementation of Anthropic's Model Context Protocol that connects your local development environment (like Cursor or Claude Desktop) to Bolti's backend APIs.

Instead of jumping between a browser dashboard, terminal windows, and your codebase, you can ask the LLM in your editor to perform complex actions on your Bolti workspace. The server registers approximately 40 tools across several key surfaces:

• Agents: List, create, update, or delete agents, and manage configuration options like LLMs, voices, and languages.
• Tools: Create and assign workspace HTTP tools (such as database lookups or CRM updates) directly to your agents.
• Calls & Conversations: Browse call history, retrieve conversation transcripts, and grab signed recording playback URLs.
• Phone Numbers & SIP: List assigned DIDs, register SIP trunks, and place outbound test calls over PSTN.
• Settings: Monitor workspace credit balances and update organization configurations.

How to configure the Bolti MCP server in Cursor

Setting up the Bolti MCP server inside Cursor takes less than two minutes. Because the MCP server is treated as a core control plane, you can configure it to run locally and authenticate with a personal access token.

Step 1: Generate your API token

Log into your Bolti dashboard and generate an MCP Personal Access Token from your developer settings. Keep this token handy.

Step 2: Add the server to Cursor

1. Open Cursor and navigate to Settings > Features > MCP.
2. Click on + Add New MCP Server.
3. Fill in the configuration details:
   • Name: bolti
   • Type: command
   • Command: npx -y @bolti/mcp-server (or the corresponding command provided in your setup panel)
4. Add your API token as an environment variable (e.g., BOLTI_API_KEY=your_token_here).
5. Save the configuration and verify that the status indicator turns green.

Building a voice workflow inside Cursor

Once the server is connected, the LLM in your Cursor Composer or Chat pane can interact with Bolti on your behalf. Here is how you can use natural language to build and test a voice workflow in real-time.

1. Define and configure your voice agent

You can ask Cursor to build a new outbound agent and select its voice and LLM providers. For example, you can choose a low-latency model like Groq Llama-3 or Gemini 2 Flash, and pair it with a specialized text-to-speech provider like Cartesia or ElevenLabs.

Ask Cursor:

"Create a new outbound sales agent called 'Lead Qualifier'. Set the LLM provider to Groq with the Llama-3 model, and set the voice to the Cartesia 'Blake' voice."

The LLM will call the create_agent tool and return a confirmation of your new agent's setup.

2. Create and attach an API tool

Voice agents are most useful when they can take action during a call. If you need your agent to check order statuses, you can define a tool and assign it to the agent directly from your editor.

Ask Cursor:

"Create a tool called lookup_order pointing at https://api.mycompany.com/orders/{id}. Assign this tool to my 'Lead Qualifier' agent and run a test mock call."

3. Trigger a test call and read the transcript

To test how your agent handles live interruptions, background noise, and tool calling, you can trigger an outbound call directly to your phone.

Ask Cursor:

"Place an outbound call from my 'Lead Qualifier' agent to my phone number at +919876543210."

Once the call ends, you can inspect the performance without leaving Cursor:

"Pull the transcript of the last call and highlight any parts where the agent called the lookup_order tool."

This tight feedback loop is why developers prefer Bolti over platforms like Bolna AI or Ringg AI. You can iterate on system prompts, tweak voice characteristics, and test API integrations entirely from your keyboard.

Why developers prototype voice workflows with MCP

Traditional voice development requires constantly switching contexts. You write an API endpoint in your editor, deploy it, open a browser dashboard to update the webhook URL, trigger a call using a curl command, and then search through dashboard logs to see what failed.

With the Bolti voice AI mcp server, your editor becomes the cockpit. It is particularly powerful for:

• Agent iteration loops: You can say "show me the system prompt, make the greeting warmer and shorter, and save it."
• One-off operational tasks: You can instruct the model to "rotate the API key on the database tool for all 5 active support agents in this workspace."
• Debugging: Quickly pull down conversation logs and transcripts to pinpoint exactly why an LLM failed to trigger a tool call.

If you want to see how other engineering teams are utilizing these automated workflows to handle thousands of concurrent support calls, you can read our case studies.

Set up your first voice agent in Cursor

Prototyping automated phone agents no longer requires complex telephony integration or tedious dashboard configuration. By combining Cursor with Bolti's native MCP server, you can build, test, and refine high-performance voice agents using natural language.

Spin up your first voice-enabled workflow in under 10 minutes. You can create your developer account to get 50 free minutes of call time, or deploy production-ready lines immediately using our flat ₹6/minute pay-as-you-go pricing.

Frequently Asked Questions

What is the Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open standard developed by Anthropic that allows local LLM clients like Cursor or Claude Desktop to securely connect with external tool servers, enabling the LLM to read data and execute actions on third-party platforms.

Does the Bolti MCP server cost extra to use?

No, the Bolti MCP server is completely free and open-source. You only pay for the standard usage of the platform, which is priced at a flat rate of ₹6/minute for active voice calls.

Can I use the Bolti MCP server with Claude Desktop?

Yes, the Bolti MCP server is fully compatible with any MCP-compliant client, including Claude Desktop, Cursor, and VS Code extensions that support the protocol.

What security controls exist for the MCP server?

The MCP server runs locally on your machine and communicates with Bolti using personal access tokens that you generate. You must manually approve any tool execution or state change proposed by the LLM in your editor.