Bland

What it does

Writes code

NoNorm turns a plain-English description into a conversation pathway, persona, voice and tools; nothing on the site describes it authoring application source code.Plain English in. Pathway, persona, voice, and tools out. You decide what to add next.bland.ai/product ↗

Debugs and fixes

PartlyNorm reproduces a failing call, finds the root cause and applies the fix, but the fix lands in a Bland pathway, not in your codebase.Norm pulls in the failing call and its pathway, reproduces the failure, applies a fix in a draft, and verifies it against the original call.docs.bland.ai/tutorials/norm ↗

Writes tests

PartlyIt generates and runs call-scenario suites with assertions against the agent, including scenarios derived from historical transcripts; these are conversation tests, not unit tests for code.Scenario suites with assertions. Back-test against historical calls. Run thousands in parallel.bland.ai/product ↗

Reviews code

NoReview runs the other direction: a human reviews Norm's draft before it merges. No surface reviews a diff someone else wrote.Every change Norm makes is reviewable on the issue, so you can see exactly what it touched before merging the fix into production.docs.bland.ai/tutorials/norm ↗

Whole-codebase work

NoThe unit of work is a pathway YAML directory synced to one Bland account via push, pull, diff and watch; there is no repository-wide code migration.Create and manage conversational pathways as YAML files. Sync local YAML with the Bland API, run interactive tests, and automate test cases.docs.bland.ai/sdks/cli ↗

Ships it

PartlyIt ships agent versions to live phone traffic with canary, promotion and rollback; it does not deploy application code.Canary new versions against 1% of real calls. Promote, or roll back in seconds. Zero-downtime deploys, every time.bland.ai/product ↗

Done-for-you service

PartlyYou run the software yourself on the self-serve plans, but Enterprise buyers get Bland engineers building the first deployment.Our forward-deployed engineering team works directly with yours to build, optimize, and scale the deployment. Enterprise plans include FDE support.bland.ai/pricing ↗

How much it does unattended

Runs autonomously

YesAgents run whole phone calls unattended at volume; a human only re-enters when a rail escalates or transfers.Bland automates the calls you thought required a human. Running entirely on your own infrastructure for maximum security.bland.ai ↗

Multi-agent

PartlyWarm transfer runs a second proxy-agent leg alongside the primary agent, but that is call handoff, not general orchestration.When a transfer is needed, the agent initiates a second proxy agent call in the background.docs.bland.ai/platform/billing ↗

Agent permissions

PartlyGuard rails bound what an agent may say and can end, transfer or reroute a call; custom rails are Enterprise-only.Guard rails continuously analyze every AI and user response during production calls. They catch severe failure modes like discrimination, self-harm…docs.bland.ai/tutorials/guard-rails ↗

Which models it runs on

Claude

NoClaude never runs the call: Bland states no third-party AI provider is in the call path at all.Most voice AI vendors list OpenAI, Anthropic, ElevenLabs, and a telephony reseller as sub-processors handling your call data. Bland runs its own…bland.ai/trust-security ↗

GPT

NoOpenAI models are named as exactly what Bland does not use; inference stays on Bland's own speech and language models.Where other vendors wrap third parties, Bland runs its own model and speech.bland.ai/trust-security ↗

Open models

NoNo open-weight option is offered; every model in the call path is Bland's own, purpose-built for telephony.With every model custom-made for phone calls, your data never goes through third parties. No sudden changes to your model, pricing, or terms of…bland.ai ↗

Model choice

NoThe vendor picks the model and sells that as the differentiator; buyers get no model selection on any tier.Bland's per-minute rate covers the LLM, STT, and TTS in one number, with no token charges and no model-provider pass-throughs.bland.ai/pricing ↗

Where you use it

In your editor

YesAn MCP server puts agent building inside Claude Code, Cursor, VS Code and Codex; the agent itself still runs on calls.The same MCP server makes Bland reachable from Claude Code, Cursor, and any other AI editor your team already uses.bland.ai/product ↗

On the command line

YesA published npm CLI drives calls, pathways, numbers, guard rails, evals and releases from a terminal.Manage your entire Bland account from the terminal: make calls, build and test pathways, configure phone numbers, and more.docs.bland.ai/sdks/cli ↗

In a browser

YesA hosted dashboard covers call logs, transcripts, recordings and review; agents also embed as a web voice/chat widget.Call logs are the central hub for reviewing completed and active calls. Every call made through Bland — outbound, inbound, voice, web, or SMS —…docs.bland.ai/tutorials/call-logs ↗

Not established: In your pipeline

Whose machine it runs on

Self-hosted

PartlyVPC, on-premise and air-gapped deployments exist but the pricing table gates them to Enterprise contracts only.Self host Bland on your infrastructure (for Enterprise).bland.ai/product ↗

Open source

NoThe platform is proprietary under the Terms. Only client tooling is published: bland-cli on npm declares an MIT licence and ships a build with no public repository named.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/terms ↗

What it costs to run

How it meters

YesBilling is per connected minute, prorated to the second, with a separate transfer-minute rate rather than seats or tokens.Prorated to the exact second. Only active call time is counted.docs.bland.ai/platform/billing ↗

Free tier

YesYou can sign up and build without a sales call or a card, though call minutes are billed once the two credits are gone.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/pricing ↗

API access

YesA documented public REST API covers calls, pathways, knowledge bases, numbers and webhooks on every plan.Hit the REST API from your terminal, your code, or the CLI.bland.ai/product ↗

MCP server

YesBland hosts a remote MCP server so another agent can place calls, read transcripts and manage pathways with an API key.It's a remote Model Context Protocol server hosted by Bland. There's nothing to run locally: point your client at the endpoint, authenticate with…docs.bland.ai/integrations/mcp/overview ↗

Bring your own key

PartlyYou can bring your own telephony and, on Enterprise, your own cloud or hardware, but never your own model keys.Yes. Bland integrates with your existing Twilio account, with SIP trunks from any contact-center or telephony provider, or you can use Bland's…bland.ai/pricing ↗

Buying it for a team

A company can buy it

YesAccounts are organisation-scoped with enterprise workspaces, contracted volume pricing and admin-managed identity.Bland supports both OIDC (OpenID Connect) and SAML 2.0 based Single Sign-On (SSO) for enterprise workspaces.docs.bland.ai/enterprise-features/SSO ↗

Seat model

NoNothing is priced per user: plans charge a flat platform fee plus usage, and the comparison table meters calls, concurrency, voices and knowledge bases rather than seats.$299/month platform feebland.ai/pricing ↗

Pooled budget

YesCredits and limits sit on the organisation, so everyone with a key draws from one shared balance and one set of caps.This one-time credit has been automatically applied to your organization and will support your upcoming usage. Each organization's credit is…docs.bland.ai/platform/billing ↗

Admin controls

PartlyAdmins constrain agent behaviour with time-based and continuous rails plus role-based access, but only Enterprise gets custom rails.Enterprise customers can create up to 5 custom guard rails for industry-specific concerns/company policies.docs.bland.ai/tutorials/guard-rails ↗

Audit log

PartlyCall-level auditing is thorough — every rail trigger and disclosure is logged — but no admin-action audit trail is documented.Audit trail Tamper-resistant call logs. Every disclosure recorded, every rail logged.bland.ai/product ↗

Single sign-on

PartlyBoth OIDC and SAML 2.0 are supported with auto-provisioning, but the plan table shows SSO only in the Enterprise column.Bland supports both OIDC (OpenID Connect) and SAML 2.0 based Single Sign-On (SSO) for enterprise workspaces.docs.bland.ai/enterprise-features/SSO ↗

What happens to your code

Opt out of training

PartlyNo outside model provider sees your calls and zero-retention processing exists for sensitive work, but no general training opt-out is published.Zero-retention processing available for sensitive workflowsbland.ai/trust-security ↗

Data residency

PartlyUS, EU and APAC regions with per-call pinning are offered, but the plan table restricts contractual data residency to Enterprise.Data residency US, EU, and APAC deployment options on request.bland.ai ↗

Getting out

PartlyCall data exports as CSV and via API, but the export window is capped at 180 days and credit expiry is never addressed.Click Export Calls in the header to export your call data as a CSV delivered via email.docs.bland.ai/tutorials/call-logs ↗

Certifications

YesSOC 2 Type II, HIPAA with BAA, GDPR and PCI DSS v4.0 are all claimed, with reports released under NDA on Enterprise.Yes. SOC 2 Type I and Type II, HIPAA-eligible with a signed BAA, GDPR, and PCI DSS. Compliance documentation is available under NDA on Enterprise.bland.ai/pricing ↗