Le Répertoire des AgentsFR · EN · ES · JA · HI

“Nobody picks up my cold calls, and I'm wasting my time”

People who have to qualify a lot of prospects by phone: swamped switchboards, wrong numbers, voicemail. Spending your days on the phone to talk to three people is discouraging.

Recommended agent: Bland

💰 Usage-based pricing, depending on call volume (check the rates on bland.ai). Level: Intermediate 🌍 United States Free trial credits at signup (no card), then usage at $0.14/min; no permanent free plan

Bland (bland.ai), AI voice agents that make your calls for you. They dial the numbers, converse naturally, qualify the prospect with your questions and only transfer the promising calls to you — or book the meeting directly.

How it works

Trigger: You launch a list of calls to qualify — dials the numbers, converses and asks your questions, and you get booked meetings or hot calls transferred to you.

Step-by-step setup

Create a Bland account and write the agent's qualification script.
Pick a voice and test the agent by calling it yourself.
Import your list of numbers to qualify.
Launch the campaign and get the hot prospects in your CRM.

⚠️ Guardrails: Test the agent yourself before unleashing it on real prospects: a badly tuned voice scares people off. Let callees know they're speaking with an AI assistant and follow the rules on telemarketing.

🔧 Technical details

Installation: hybrid — cloud platform accessible via dashboard and API (self-serve plans), with dedicated on-premise or VPC deployments in Enterprise.
Prerequisites: A Bland account and an organization API key; a phone number (an inbound number is included in the Start plan, or your own telephony via Twilio or a SIP trunk). No card required to start the Start plan.
Hardware: non_applicable
Self-hosting: yes — documented self-hosted and on-premise deployments for regulated industries, and an “On-Prem/VPC” option in the Enterprise plan (pricing page); the full voice stack then runs on the customer's infrastructure.
API & integrations: API: yes — REST API (v1 for calls and post-call webhooks, v2 for agents) on api.bland.ai, authenticated with an organization API key. · SDK: yes — Web Agent SDK to embed a voice agent in a web application (token authentication), and Bland TTS toolkit (Node library, CLI and MCP server). · CLI: yes — official CLI to manage the account from the terminal: calls, building and testing pathways, number configuration. · Webhooks: yes — post-call webhooks (call data sent to the customer's server, with an HMAC signature to verify) and webhook steps in pathways to call external APIs during the call. · MCP: yes — official Bland MCP server (connection to code agents and personal agents, “Sign in with Bland” OAuth) and a Norm plugin to build Bland agents from a code agent.
Models: Bland's proprietary models for the whole chain (speech recognition, language model, speech synthesis): no third-party model provider is involved, and the per-minute rate includes LLM, STT and TTS with no token fees. No documented model BYOK; on the other hand, you can bring your own telephony (existing Twilio or another carrier's SIP trunk).
Data: Bland runs its entire voice stack (telephony, speech recognition, language model, speech synthesis) on its own infrastructure: call data passes through no third-party AI provider, and a single BAA covers the whole call in healthcare. Encrypted data, configurable retention and audit logs on every call (official healthcare page); in on-premise/VPC deployments, data stays in infrastructure controlled by the customer. SOC 2 Type I and II, HIPAA, GDPR (DPA) and PCI DSS announced on the pricing page.
Documented limits: According to the official pricing page: Start plan — 10 concurrent calls, 100 calls/day and 100 calls/hour, 10 knowledge bases, 1 voice; Build plan — 50 concurrent calls, 2,000 calls/day, 1,000 calls/hour, 50 knowledge bases, 5 voices; Enterprise — custom concurrency and volumes, unlimited caps. Announced uptime SLA: 99.9%.
Operations: Dev, staging and production environments with pinned versions and controlled promotion; evaluations and test scenarios, guardrails and supervision alerts; in Enterprise: SSO (OIDC or SAML 2.0), JWT signatures for webhooks and requests, data residency, dedicated infrastructure and a dedicated engineer (FDE).

Not documented in the official sources: License.

Official sources: docs.bland.ai · www.bland.ai · www.bland.ai · docs.bland.ai · docs.bland.ai — verified on November 1, 2026.

Find your agent with the quiz →

Frequently asked questions

How much does Bland cost?

Usage-based pricing, depending on call volume (check the rates on bland.ai). Free trial credits at signup (no card required), then usage at $0.14/min; no permanent free plan.

Is Bland free?

No permanent free plan, but the trial is free. Free trial credits at signup (no card required), then usage at $0.14/min; no permanent free plan.

Is Bland hard to get started with?

Level: Intermediate. Setup takes 4 steps; the first: Create a Bland account and write the agent's qualification script.

What limits and guardrails should I know about with Bland?

Test the agent yourself before unleashing it on real prospects: a badly tuned voice drives people away. Inform callees that they are speaking with an AI assistant and follow the rules on telemarketing calls.