Bolna
Create conversational voice agents using Bolna AI to enhance interactions, streamline operations and automate support.
Verdict
Common use cases
- Trigger outbound sales calls from Slack threads
- Pull call transcripts into project briefs
- Monitor agent performance across campaigns
- Delete test batches after QA runs
- List available phone numbers before launch
Integration
- Vendor
- Bolna
- Category
- other
- Auth
- API_KEY
- Tools
- 15
- Composio slug
bolna
Tools
- Delete agent by iddestructive
Delete an agent
- Delete batch by iddestructive
Delete a batch
- Fetch all batches by agent id
List all batches for agent
- Get all agents
List all agents
- Get all phone numbers
List all phone numbers for your account
- Get execution by id
Retrieve specific execution by `execution id`
- Make a phone call from agent
Initiate calls
- Retrieve agent by id
Retrieve an agent
- Retrieve agent execution details
Retrieve specific execution by an agent
- Retrieve agent execution status
Retrieve all executions by an agent
- Retrieve batch details by id
Retrieve a batch
- Retrieve batch execution list
Retrieve all executions from a batch
- Schedule batch by id
Schedule a batch for calling via agent
- Setup inbound call for agent
Add agent for inbound calls
- Stop batch by id
Stop a running batch
Setup
Setup guide
- 11. Open your Switchy workspace and navigate to Settings > Integrations > Browse MCPs. 2. Search for Bolna and click Connect. 3. Log into your Bolna account, go to Settings > API Keys, and generate a new key with full read-write access. 4. Paste the key into Switchy's prompt and click Authorize. 5. Return to any Space and type '@Bolna get all agents' to confirm the connection works—you should see a list of your configured voice agents. 6. To test a real workflow, create a new Space, @mention Bolna, and ask it to retrieve execution details for a recent call using 'Get execution by id' with a valid execution ID from your Bolna dashboard.
What teammates see: by default, memories from Bolna are scoped to the Space (PROJECT visibility) - you can mark any memory PRIVATE or share it ORG-wide.
Works well with
Top models
Compatibility data appears once enough Spaces have used this MCP together with a given model.
How Switchy teams use it
Starter prompts
List Active Agents
@Bolna get all agents and show me their names, IDs, and current status in a tableOpen in a Space →
Start Outbound Campaign
@Bolna make a phone call from agent ID [agent_id] to +1234567890 and log the execution ID hereOpen in a Space →
Review Call Execution
@Bolna get execution by id [execution_id] and summarize what happened on the callOpen in a Space →
Check Batch Progress
@Bolna fetch all batches by agent id [agent_id] and tell me completion rates for each batchOpen in a Space →
Inventory Phone Numbers
@Bolna get all phone numbers and show me which ones are currently assigned to agentsOpen in a Space →
Example outputs
Illustrative - representative of the model's voice and quality, not literal recordings.
@Bolna list all my voice agents and show me which ones have active phone numbers assigned
Found 4 agents in your Bolna account. Two are production-ready: 'Customer Support Bot' is assigned to +1-555-0198 and has handled 127 calls this month. 'Appointment Scheduler' uses +1-555-0243 with 89 calls logged. The other two agents ('Sales Qualifier' and 'Feedback Collector') are in draft status with no phone numbers assigned yet. All agents use your default voice model configuration.
This example combines the 'Get all agents' and 'Get all phone numbers' tools to give you deployment visibility. Useful for auditing which voice agents are live versus in-progress. Requires API key with read access to both agents and phone number resources.
@Bolna trigger an outbound call from my 'Appointment Scheduler' agent to +1-555-0176 to confirm tomorrow's 2pm slot
Call initiated successfully. Execution ID: exec_8k2mq9. Your 'Appointment Scheduler' agent is now dialing +1-555-0176 from your assigned number +1-555-0243. The call will follow the agent's configured conversation flow for appointment confirmations. You can monitor this execution in real-time through the Bolna dashboard or retrieve its status later using the execution ID provided.
This demonstrates the 'Make a phone call from agent' action — a write operation with immediate side-effects. The call happens in real-time and will appear on your phone bill. Ensure your agent's prompt and voice settings are tested before triggering production calls. Rate limits apply based on your Bolna plan tier.
@Bolna pull the last 20 call executions for my 'Customer Support Bot' agent and summarize common customer issues mentioned
Retrieved 20 recent executions for 'Customer Support Bot' (agent ID: agt_7x3p). Analyzing call transcripts reveals three recurring themes: 8 callers asked about order tracking delays, 5 requested password resets, and 4 inquired about return policies. Average call duration was 3m 42s. Three calls escalated to human handoff due to billing disputes. Two executions failed due to caller hangups within the first 15 seconds.
This pairs Bolna's 'Fetch all batches by agent id' or execution retrieval with the AI's summarization capabilities. Illustrative output assumes transcript access is enabled in your Bolna plan. Actual analysis depth depends on what metadata Bolna returns per execution — some plans may only provide call duration and status, not full transcripts.
Use-case deep-dives
When Bolna makes sense for high-volume cold outreach
A 6-person sales team running 200+ outbound calls per week can use Bolna to automate initial qualification conversations. The MCP exposes agent creation, batch execution, and phone number management—enough to spin up campaigns from Switchy without opening Bolna's dashboard. This works if your team already writes call scripts in-house and needs programmatic control over when agents dial. The trade-off: Bolna is a voice-AI platform, not a general telephony API, so you're locked into their agent model. If you're just logging call outcomes or pulling transcripts into your CRM, a lighter webhook integration beats this MCP. Use Bolna's MCP when your workflow is 'draft script in Switchy, deploy agent, monitor batch status'—all in one workspace.
Why this MCP is overkill for most support teams
A 3-person support team handling 40 callbacks per day might think Bolna's MCP lets them automate 'we'll call you back' promises. It does—but only if you're building voice agents that conduct the callback conversation, not just dialing a number and handing off to a human. The MCP's 15 tools are agent-lifecycle commands (create, delete, batch-execute), not simple dial-and-transfer primitives. If your support flow is 'customer requests callback, agent dials, human picks up', you don't need Bolna's agent orchestration. This MCP wins when the callback itself is scripted and repeatable—think appointment reminders or order confirmations where the AI completes the interaction. For ad-hoc human callbacks, stick with Twilio or your existing phone system.
When Bolna bridges marketing automation and voice
A 4-person growth team running Facebook lead ads wants to call every lead within 5 minutes of form submission. Bolna's MCP lets them wire Switchy to trigger an agent call as soon as a lead hits their CRM, using the 'Make a phone call from agent' tool with dynamic variables from the lead record. The agent asks qualifying questions, logs responses, and routes hot leads to a human closer. This setup works if your team has 100+ leads per week and a repeatable qualification script—Bolna's batch and execution tools let you monitor call outcomes without leaving Switchy. The threshold: if your close rate depends on nuanced human conversation in the first call, the agent model breaks down. Use this MCP when speed and script consistency beat personalization, and your team can iterate on agent prompts weekly.
Frequently asked
What does the Bolna MCP do in Switchy?
The Bolna MCP lets your team manage voice AI agents and phone calls directly from Switchy. You can create and delete agents, initiate outbound calls, retrieve call execution logs, and manage phone numbers across your Bolna account. It's built for teams running conversational AI workflows who want to trigger and monitor voice interactions without switching to the Bolna dashboard.
Do I need admin access to connect Bolna MCP?
You need a Bolna API key with permissions to manage agents and phone numbers. Bolna uses API_KEY authentication, so whoever connects it must have access to generate or retrieve keys from your Bolna account settings. If your Bolna workspace restricts API key creation to admins, you'll need admin involvement to set this up.
Can the Bolna MCP listen to live calls or transcribe audio?
No. The MCP handles agent configuration and call initiation, not real-time audio processing. You can retrieve execution records after a call completes, but live transcription or call monitoring happens inside Bolna's infrastructure. If you need real-time call data in Switchy, you'll need to poll execution endpoints or use Bolna's webhooks separately.
How is this different from using Bolna's API directly?
The MCP wraps Bolna's API so your AI agents can call it without you writing integration code. If you're already using Bolna's REST API in scripts, the MCP won't change your workflow. But if you want Claude or another LLM to trigger calls or fetch agent data based on conversation context, the MCP makes that possible without custom middleware.
Who on the team should connect the Bolna MCP?
Whoever manages your Bolna voice agents and has API key access. This is typically a product or ops person running conversational AI campaigns, not a general team member. Once connected, any Switchy user in your workspace can invoke the tools, so limit access if you don't want everyone initiating phone calls from your Bolna account.