At a glance
- Category
- Voice AI agents
- Tags
- GDPR, HIPAA, Private AI, Self-hosted, SOC 2
For developers
- Works with
- Salesforce, Nice, Vonage, Slack, HubSpot, Twilio and 6 more
Bland AI in our research
All researchStudy
AI Tool Pricing Index, November 2026
Pricing
Usage-based
43 of the 361 tools we checked bill by usage.
Edition price, checked 28 October 2026.
Pricing pages of the AI tools, November 2026 editionas of n = 361 tools checked
Bland runs phone agents on a stack it owns end to end
Bland is an AI phone agent platform: agents take inbound calls, place outbound ones, and are described by the vendor as working across phone, SMS, and web. The pitch that separates it from Vapi and Retell is ownership of the layers. Bland says it operates its own speech-to-text, language model and text-to-speech rather than wiring third-party vendors together, and states on the homepage that it is "The only voice AI that runs entirely on its own models".
That claim has a commercial consequence and a control consequence. Commercially, one per-minute rate covers the model, transcription, synthesis and telephony. For control, the vendor argues that because calls do not leave to a third party, customers are "never victim to unexpected model changes". The flip side is that you do not pick the model: no page on the site or in the documentation offers a model selector, an open-weight option, or a way to supply your own provider key. The only bring-your-own path documented is telephony, where Bland connects to your existing Twilio account or SIP trunks.
The vendor's legal name is Intelliga Corporation, trading as Bland AI, governed by California law.
Building an agent is prompt-first, with pathways added where needed
An agent in Bland starts as a prompt rather than a flowchart. The builder's guidance is to "Write in plain language" and add structure only when the conversation requires it. Two prompt fields stack: one scoped to the current node, one applied to every node for tone. Tools and knowledge sources are attached by typing @ in the prompt, and the documentation warns that building a tool without tagging it fails silently.
Where order matters, a Pathway gives a canvas of steps with custom code and webhook nodes, and an Authentication zone gates everything placed beneath it behind identity questions. Guardrails sit on a second tab and are checked live on every call, with optional actions to end, transfer or redirect. Four built-in guardrails cover TCPA disclosures and are available on every plan; custom guardrails require an enterprise plan.
Shipping is two steps that catch people out. Publish moves an agent from Development to Staging, Promote moves Staging to Production, and the docs state plainly that "An agent that has not been promoted to Production will not answer calls".
A platform fee plus metered minutes, and tiers that disagree
The entry plan, Start, carries "$0 platform fee. No card required" with two credits and an inbound number, then charges $0.14 per connected minute. Build is $299 a month at $0.12/min. Transfer minutes are billed separately at $0.05 and $0.04, outbound attempts carry a $0.015 minimum, SMS is $0.02 per message, and web widget messages are $0.01 each. Norm, the building assistant, is billed on its own token-based model.
The two pages do not agree on what you can buy. The pricing page lists three plans, Start, Build and Enterprise. The billing documentation lists five, adding a Scale plan at $499 with $0.11/min and the Agent Phone Plan at $29.99 a month. Scale is absent from the public pricing table entirely, which matters because it is the cheapest route to 100 concurrent calls.
The Agent Phone Plan is the odd one out: a single number with unlimited US and Canada voice and SMS in place of per-minute billing, $14.99 for the first month then $29.99, explicitly scoped to one bot doing one thing at a time and unavailable to Enterprise accounts.
Nearly everything a contact center needs sits behind Enterprise
Read the plan comparison downward and the self-serve tiers thin out fast. SMS and web chat, warm transfers, live transfers, the appointment scheduling node, guardrails as protected calls, alarms and monitoring, citations, outcomes, custom dialing and custom code extraction all carry a dash on Start and Build and Included on Enterprise. The compliance block is the same shape: BAA, SSO, JWT signatures and data residency are Enterprise rows.
Enterprise also changes the infrastructure underneath. Each enterprise organization runs on dedicated Kubernetes pods rather than shared capacity, chooses when to adopt a new runtime release, and can run canary deployments that route a percentage of real calls to a new version before promoting it.
No price is published for any of it: the Enterprise column reads Custom, contracted to volume.
The API and MCP server are first-class, not an afterthought
The REST API is documented in depth: POST /v1/calls dispatches a call and returns a call ID immediately, a webhook or a poll on the same endpoint returns status, transcript and recording, and webhook payloads can be verified with HMAC signatures.
Bland also hosts a remote MCP server at api.bland.ai/v1/mcp, reachable with an API key or OAuth sign-in, exposing call placement, pathway validation, agent lifecycle, analytics and evals, with a documented rate limit of 120 requests per minute per organization. Mutating tools are annotated destructive and require an explicit confirmation argument. There is an onboarding flow written for bots with no account at all, where the owner approves in a browser or relays an SMS code and the agent receives its own key.
What is published on security, and what is not
Bland claims SOC 2 Type I and Type II, HIPAA with a signed BAA, GDPR with a DPA and PCI DSS self-attestation, and adds FedRAMP 20x Class A and ISO 27001 on the same FAQ answer. Documentation is offered under NDA on Enterprise, and a Vanta-hosted trust center is linked from the homepage. A 99.9% uptime SLA appears in the comparison table for every tier, not only for contracts.
Two things on the data side do not line up neatly. The homepage says customer data "never trains other models", which is a statement about third parties rather than about Bland's own training. The terms of service separately grant Bland a "perpetual, worldwide, non-exclusive, royalty-free, irrevocable license" over user content in connection with operating the service. The privacy policy sets no retention period, saying only that information is kept as long as necessary.
Per-agent controls exist for the obvious risks: recording can be disabled under advanced settings for PCI, responses can be excluded from logs, and memory, which writes facts into a contact record, is on by default in a new agent.
Where it fits, and where it does not
The buyer Bland writes for is a regulated operation with real call volume: healthcare, insurance, financial services and logistics, replacing an IVR first and expanding from there. The customer stories on the site are in that shape, including an IVR retirement at Idaho Housing reported as $750K in annual savings. Outcome figures throughout are vendor-published customer quotes rather than independently audited results.
It fits poorly for a small team that needs transfers, SMS or scheduling, since those are Enterprise rows rather than self-serve features. It also fits poorly where cost predictability matters more than call quality: the bill is minutes, and nothing on the pricing page estimates a month for you.
Bland AI price and terms
Terms
- Free to start
- PartlyLimited free use
- Card to start
- YesNot required
- Sign up without sales
- YesYes
The terms in full
Signing up on Start costs no platform fee and needs no card, but the free usage is a one-off grant of 2 credits plus an inbound number the page values at $15/month; calls beyond those credits are billed at $0.14/min, and Start is capped at 100 calls/day, 10 concurrent calls, 1 voice and 10 knowledge bases.
For AI agents
- MCP server
- Yesdocs.bland.ai,
- API
- Yesdocs.bland.ai,

Pricing page, captured
bland.ai/pricing, checked
All plans and limits
| Plan | Price | Concurrent calls | Daily calls | Hourly calls |
|---|---|---|---|---|
| Start | $0 a month | 10 calls | 100 calls/day | 100 calls |
| Build | $299 a month | 50 calls | 2,000 calls/day | 1,000 calls |
| Enterprise | Price on request | Concurrency sized to your volume | Unlimited | Unlimited |
Read from bland.ai/pricing.
What we checked on Bland AI
What the price counts
Billing is metered per minute of call time, $0.14/min on Start with no platform fee and $0.12/min on Build behind a $299/month platform fee, with transfer minutes priced separately and Enterprise quoted to volume.
On the page: 'Compare plans' table, Usage + scale section, Talk Time Rate and Transfer Time Rate rows
bland.ai/pricingChecked
Free to use
You can sign up on Start without a card and without a platform fee, but the only free usage is a one-off grant of two credits plus an inbound number; past that every minute is billed at $0.14, so there is no standing no-cost plan.
Sign up on the Start plan and you get two credits plus an inbound number valued at $15/month, with no card required.
bland.ai/pricingChecked
Public API
A documented public REST API places calls, polls their state and returns transcripts, authenticated with a bearer API key from the dashboard.
POST /v1/calls dispatches the call and returns immediately with a call_id. It does not wait for the call to finish.
MCP server
Bland hosts a remote MCP server at api.bland.ai/v1/mcp with OAuth or API key auth, covering calls, pathways, analytics and evals.
The Bland MCP server lets any AI agent work with your Bland account through natural language: place and inspect calls, send texts, build and validate…
Open source
The platform is proprietary: the terms claim copyright over all software, code and service materials, and no public repository or license is offered.
All Materials are the copyrighted property of Bland AI or its subsidiaries or affiliates under the laws of the United States and other countries.
bland.ai/legal/termsChecked
Bland AI alternatives
Bland AI and VoiceDrop
Bland AI beside its alternatives: the price on each one's site and the day it was checked Tool Price Checked Bland AI From $0.14 per minute VoiceDrop From $95 a month
Questions about Bland AI
- How much does Bland AI cost?
- Bland AI costs $0.14 per minute. From bland.ai/pricing, checked 28 October 2026.
- Does Bland AI have an API or an MCP server?
- Bland AI has an API. From docs.bland.ai/platform/agent-quickstart, checked 20 October 2026. Bland AI has an MCP server. From docs.bland.ai/integrations/mcp/overview, checked 20 October 2026.
- Is Bland AI open source, and can you host it yourself?
- Bland AI is not open source. From bland.ai/legal/terms, checked 20 October 2026. You can host Bland AI yourself on Enterprise. From bland.ai, checked 20 October 2026.
- What have we checked on Bland AI?
- 5 facts, under these headings: What it costs to start, how you build with it.
- How do we check Bland AI?
- We hold 5 facts on Bland AI, quoted from 12 pages. Facts and price last checked 28 October 2026, when we also found the last change.
Sources and changes
Last changed
Sources12
- bland.ai/pricing
- bland.ai/
- docs.bland.ai/agents/agent-builder
- bland.ai/trust-security
- docs.bland.ai/agents/conversations
- docs.bland.ai/platform/agent-phone-plan
- docs.bland.ai/tutorials/post-call-webhooks
- docs.bland.ai/enterprise-features/SSO
- docs.bland.ai/integrations/mcp/overview
- bland.ai/legal/terms
- docs.bland.ai/platform/agent-quickstart
- docs.bland.ai/platform/billing
What changed1
- Free tier
Bland AI price history
We have read the price once so far, on 28 October 2026: from $0.14 per minute.
| Price | Plan | First seen | Last seen | Readings | Source |
|---|---|---|---|---|---|
| $0.14 per minute | 28 October 2026 | 28 October 2026 | 1 | Pricing page, read 28 October 2026 |
From our research on 28 October 2026. A date is when we read the page, not when the price changed.
Is Bland AI yours?
Free, and the page then carries the verified owner mark. You confirm it with an email at bland.ai. Claim this listing
Report an error if a fact is wrong.