Bland AI: Voice Agent API, Features, and Pricing (2026)

Bland AI is an API-first platform for building AI voice agents that make and receive phone calls. It combines the pieces a production phone agent needs—telephony, speech recognition, text-to-speech, conversation logic, live tool calls, transfers, testing, and call analytics—behind one platform.

At its simplest, you give Bland AI a phone number and a task, then receive a call ID. At the other end of the spectrum, you can build a branching conversational workflow, connect it to business systems, launch a batch campaign, transfer selected calls to people, and send structured results back to your application.

This guide explains how Bland AI works, what its main features are, what the current pricing actually includes, and when it is a sensible choice for a voice automation project.

What Is Bland AI?

Bland AI provides infrastructure for inbound and outbound AI phone calls. Developers can control it through a REST API, while operations teams can design call behavior in a visual editor called Conversational Pathways.

Common use cases include:

• Qualifying or following up with leads
• Confirming, rescheduling, and reminding people about appointments
• Handling routine customer service calls
• Collecting structured information by phone
• Routing callers to the correct human team
• Running surveys and outbound notification campaigns

Unlike a basic voice chatbot, a Bland agent can take actions during the conversation. It can look up an order, schedule a meeting, write data to a CRM, send a Slack message, or call a custom API. After the call, it can return the transcript, summary, extracted variables, disposition, duration, and other metadata through a webhook.

How Bland AI Works

A typical Bland AI workflow has six stages:

1. Define the conversation. Use a task prompt for a simple call, a reusable persona, or a Conversational Pathway for a controlled multi-step flow.
2. Choose the phone setup. Use Bland-provided telephony, import a Twilio number, or connect existing infrastructure through SIP on supported plans.
3. Start or receive a call. Create an outbound call with the API, send a batch, schedule a call, or attach an agent to an inbound number.
4. Act during the conversation. The agent can query a knowledge base, call external tools, load dynamic data, navigate phone menus, or transfer to a person.
5. Monitor and protect the call. Configure maximum duration, voicemail behavior, recording, live events, and guardrails for defined failure conditions.
6. Process the result. Fetch the call record or receive a post-call webhook with the transcript and structured outcome data.

The smallest outbound request needs only a destination and a task:

curl --request POST \
  --url https://api.bland.ai/v1/calls \
  --header "authorization: YOUR_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
    "phone_number": "+15551234567",
    "task": "Confirm the appointment time and answer scheduling questions."
  }'

The API returns a unique call ID that can be used to inspect status and results. Production requests can add a pathway, voice, language, tools, transfer rules, call metadata, webhooks, and other controls documented in the Send Call API.

Core Features

Inbound and Outbound Calls

Bland supports both directions of a phone workflow. Outbound calls can be sent immediately, scheduled, or dispatched in batches. Inbound numbers can have their own agent configuration, pathway, webhook, and request context.

For organizations that already have carrier infrastructure, Bland supports bringing a Twilio account and enterprise SIP configurations. This matters because the carrier, phone-number, geographic, and transfer requirements can be as important as the agent itself.

Conversational Pathways

For a one-purpose call, a well-written task prompt may be enough. More complex calls benefit from Conversational Pathways, Bland's graph-based workflow builder.

A pathway is made of nodes and labeled branches. A node can generate dialogue, say a fixed sentence, collect variables, query a knowledge base, call a webhook, transfer the call, or end it. The model decides which labeled branch matches the conversation, while the graph constrains the set of valid next steps.

This gives teams more control than one long prompt. A scheduling agent, for example, can have distinct branches for confirming an appointment, finding another time, handling an unknown caller, and escalating to staff.

Tools, Knowledge, and Dynamic Data

Bland agents can call tools while someone is still on the line. Built-in integrations cover actions such as scheduling, CRM updates, and team notifications, while custom tools can call an external API.

Knowledge bases supply reference material for questions, and dynamic data can load customer or account context at the beginning of a call or during the conversation. This separates stable instructions from information that changes for every caller.

Treat every write-capable tool carefully. Looking up an order and canceling it are different authority levels, even if they use the same API. Validate inputs, restrict what the endpoint can do, make important operations idempotent, and require human confirmation where mistakes would be costly.

Human Transfers

A voice agent should know when to stop automating. Bland can transfer a live call to a single number, route to different departments, or use pathway transfer nodes under specified conditions.

Design the handoff as part of the workflow rather than as an emergency afterthought. Pass the reason for transfer and collected context to the human side, and define what happens when no representative is available.

Batch Campaigns and Webhooks

Batch Calls can launch a campaign from a CSV or the API. Each recipient can have different variables, allowing the same pathway to reference a name, appointment, account state, or other per-call context.

Status webhooks report campaign progress, while post-call webhooks deliver completed-call data to a CRM, database, analytics pipeline, or follow-up workflow. Use webhook signatures where available, acknowledge deliveries quickly, and process expensive work asynchronously.

Testing, Standards, and Guardrails

Voice agents need more than a successful demo call. Callers interrupt, provide incomplete details, go off script, reach voicemail, ask for a human, or phrase the same intent in unexpected ways.

Bland's testing tools let teams simulate caller personas and check assertions against pathways. Its Testbed isolates individual pathway nodes, while Standards turn known interactions into repeatable regression checks. Running the same scenario multiple times is useful because model behavior is probabilistic rather than perfectly repeatable.

Guardrails monitor live responses for configured requirements or prohibited behavior and can end a call, transfer it, or move to another pathway node. Availability depends on the plan. Guardrails are an additional control, not a substitute for consent management, legal review, representative monitoring, or carefully limited tools.

CLI and MCP Support

Bland also ships an npm-based CLI for making calls, managing pathways, inspecting logs, and configuring phone numbers. The CLI includes an MCP server, so compatible coding agents can work with a Bland account through tools rather than copied commands.

That MCP connection is mainly an operator and developer interface: it lets an agent create calls or manage Bland resources. It should not be confused with the real-time voice agent that speaks to a caller.

Bland AI Pricing

Bland combines a plan fee with usage charges. The public prices listed on September 11, 2026 were:

| Plan | Platform fee | Connected-minute rate | Transfer-minute rate | Limits | |---|---:|---:|---:|---| | Start | $0 | $0.14 | $0.05 | 10 concurrent calls, 100 calls/day, 10 knowledge bases | | Build | $299/month | $0.12 | $0.04 | 50 concurrent calls, 2,000 calls/day, 50 knowledge bases | | Enterprise | Custom | Custom | Custom | Concurrency and volume sized to the deployment |

The Start plan includes two initial credits and an inbound number, but it is not unlimited free calling. Call time is billed to the second. Bland says its connected-minute rate includes the LLM, speech-to-text, and text-to-speech components; carrier and telephone-number costs are separate.

You can use Bland's built-in carrier setup at pass-through cost, connect your own Twilio account, or discuss SIP options for larger deployments. Bring-your-own-telephony customers do not pay Bland's transfer-minute fee, but still pay their carrier and Bland's AI usage rate.

Enterprise adds deployment and governance options such as on-premises or VPC availability, SSO, data residency, BAA support, higher call limits, and forward-deployed engineering. Verify feature availability and the latest rates on the official pricing page, because both self-serve plans and enterprise contracts can change.

The Separate Agent Phone Plan

Bland also offers a $29.99 monthly Agent Phone Plan: one US number with US and Canada calling and texting under fair-use limits. It permits one concurrent call and is designed to give a general-purpose AI agent its own line. This is a separate product from the developer plans above and is not the right pricing model for a high-volume contact-center workflow.

Bland AI vs. a Modular Voice Stack

The main architectural choice is whether to buy a bundled platform or assemble the voice pipeline yourself.

| Bland AI | Modular stack | |---|---| | One API and bill for the AI voice layer | Separate telephony, transcription, model, speech, and orchestration vendors | | Faster path to a working phone agent | More integration and operational work | | Pathways, batch calls, testing, and call logs included | Each capability can be selected or built independently | | Less control over individual model components | Greater freedom to tune or replace every component | | Vendor-specific workflow and deployment model | More portable application logic if designed carefully |

Bland is strongest when speed to production and one operational surface matter more than choosing every underlying provider. A modular stack is more attractive when a team has unusual latency, model, regional, cost, or infrastructure requirements and the engineering capacity to own them.

When Bland AI Is a Good Fit

Consider Bland AI when:

• Phone calls are a core workflow rather than an occasional experiment.
• You need both inbound and outbound agents on one platform.
• Conversations require branching logic, live business data, or actions.
• Operations teams need a visual workflow editor alongside an API.
• You plan to run batches and process structured results with webhooks.
• Human transfer, regression testing, and enterprise deployment controls matter.

It may be more platform than you need for a single low-volume notification flow. It may also be a poor fit if you require complete control over the language model, speech providers, or open-source runtime.

What to Evaluate Before Adopting It

Run a representative pilot rather than judging the platform from one scripted call:

1. Conversation quality: Test interruptions, silence, accents, background noise, spelling, numbers, and domain-specific names.
2. Latency: Measure the slowest turns and tool calls, not only the average response.
3. Task completion: Track whether the call achieved the business outcome without inventing information or skipping required steps.
4. Escalation: Verify transfers, context handoff, unavailable-agent behavior, and opt-out handling.
5. Reliability: Exercise carrier failures, webhook retries, duplicate events, tool timeouts, and provider outages.
6. Unit economics: Include connected minutes, transfers, phone numbers, carrier charges, failed attempts, and the monthly plan fee.
7. Governance: Confirm consent, disclosure, recording, retention, and calling-hour requirements with qualified counsel in every place you operate.

The Bottom Line

Bland AI turns a complicated real-time voice stack into a relatively direct product surface: define the conversation, connect business actions, attach a number, and call the API. Conversational Pathways and the testing system make it suitable for workflows that need more control than a free-form voice prompt, while batches, transfers, and webhooks cover the operational side of phone automation.

The tradeoff is the same as with any managed platform: you gain speed and an integrated toolset, but accept its pricing, workflow model, and place in your critical path. If those constraints fit, Bland AI is a serious option for building production phone agents without owning the entire voice infrastructure yourself.