An open-source conversational AI stack gives your engineers the code to host, assemble, and run themselves, while a proprietary conversational AI platform runs every layer of the call for you. On a phone call, the difference is who owns hosting, telephony, models, monitoring, and the fix when a call goes wrong.
A phone call shows the cost of that split quickly. A caller asks something off-script, changes an answer from two minutes ago, or goes quiet to find a card, and somebody has to have built the logic that copes.
Bland builds AI phone agents that answer inbound calls and place outbound calls. Your team builds and edits them in plain English with Norm or in Bland's visual builder, with no engineers needed on your side. Enterprise customers also get Bland's forward-deployed engineers.
What does open source mean for a conversational AI stack?
Open source means the code of a layer is published under a license that lets a team read it, run it, and change it. A phone agent has several layers: speech recognition, a language model, dialogue logic, text to speech, and the telephony that connects a call to all of them. For background on those layers, read what conversational AI is.
A proprietary platform keeps its code closed and runs those layers itself, so a team configures agents instead of assembling a stack. The two approaches can also meet in the middle, because an open-source project can have a hosted commercial offering.
As of November 2026, LiveKit is an open-source platform for building voice, video, and physical AI agents, and it includes the Agents framework for Python and Node.js.
LiveKit's Agents framework and media server are completely open source and available to run locally or host on your own infrastructure. LiveKit says the framework and entire LiveKit ecosystem are open source under the Apache 2.0 license, and that security teams at large companies audit the same code that runs on LiveKit Cloud. LiveKit Cloud is the hosted commercial offering.
What does a team take on when it assembles an open-source stack?
A team that assembles an open-source stack takes on hosting, telephony, model choice, monitoring, compliance evidence, and the fix when a call goes wrong. LiveKit's own comparison of running its framework itself against LiveKit Cloud shows where hosting, telephony, models, and monitoring land.
- Hosting: when a team runs LiveKit itself, it runs and scales its own agent servers. LiveKit Cloud is the alternative, where LiveKit handles scaling, lifecycle management, isolation, and upgrades.
- Telephony: a team running LiveKit itself brings its own SIP trunking provider.
- Models: a team running LiveKit itself brings its own model provider API keys. LiveKit Inference, included with LiveKit Cloud, offers models from OpenAI, Google, AssemblyAI, Deepgram, Cartesia, Inworld, and more.
- Monitoring: agent observability on a team-run LiveKit setup is custom or external, built with OpenTelemetry data hooks.
- Compliance evidence: a team that runs the code itself gathers the evidence for its own deployment. On LiveKit Cloud, LiveKit says BAAs are available for Scale and Enterprise customers handling protected health information.
- Fixing a bad call: LiveKit's framework is made for developers who build agents in code, not configuration, so a fix to a bad call is a code change.
Each item on that list is work for an engineering team, and the trade is control: a team that runs the framework itself can read and change the framework's code.
What does a proprietary conversational AI platform run for you?
A proprietary conversational AI platform runs the hosting, telephony connection, models, monitoring, and tooling as one product, so your team spends its time on the conversation. Bland's agents answer inbound calls and place outbound calls, and Bland's conversational AI platform page covers how a long call stays on track.
What these rows mean
- Open-source column: the hosting, telephony, models, and call review rows are LiveKit's Agents framework run by a team itself, from LiveKit's comparison of self-run and LiveKit Cloud. LiveKit's framework builds agents in code, not configuration, so a fix is a code change. The compliance row is general to any code a team runs itself.
- Hosting: isolated infrastructure applies to Enterprise customers.
- Telephony: SIP lets a team bring its own carrier, and phone from Bland is billed separately.
- Call review: alerts notify by email, phone, or a custom tool.
- Fixing a bad call: Norm Fix works from a call flagged as a Triage issue.
Bland's SIP setup works from a trunk that holds the connection to your SIP infrastructure, and its call logs and alerts cover review. For how Bland's agents hold a call together, see prompt-based agents vs Conversational Pathways.
When does each approach fit?
Each approach fits a different team, and the clearest test is the work a team is ready to take on.
Use an open-source framework when:
- A team has engineers who will build agents in code and maintain the framework code it runs.
- A team will run its own agent servers, bring its own SIP trunking provider, and choose its own model providers.
- A team will build its own review of every call, from transcripts to alerts.
Use Bland when:
- A team wants Bland's forward-deployed engineers to build and launch its first agent alongside it, on Enterprise.
- A team wants non-engineers to build and edit inbound and outbound agents in plain English with Norm or in Bland's visual builder.
- A team wants live calls pinned to a production version, with a passing evaluation required before a version is promoted.
Bland also connects to a team's own systems: Salesforce, HubSpot, Cal.com, Calendly, Notion, and Slack connect from Bland's dashboard, and anything else connects through Bland's REST API and webhooks. See integrations for the list.
Where does Bland sit between the two?
Bland sits on the proprietary side: Bland operates a closed-source model it developed itself. Bland self-hosts its models and uses no third-party model providers. So Bland is not a thin layer over a third-party model such as GPT or Claude.
Every agent has development, staging, and production environments, and live calls run the pinned production version. Evaluations score an agent's calls, and a passing run can be required before a version is promoted.
Every enterprise customer runs on its own isolated infrastructure, run by Bland: dedicated compute, containers, and resources, with releases the customer chooses when to adopt and canary deployments on a share of real calls. Dedicated GPUs run by Bland are available. Bland carries a 99.9% uptime SLA, and Bland has 250+ enterprise customers.
On compliance, Bland is SOC 2 Type II, HIPAA compliant with a BAA, GDPR compliant with a DPA, and PCI DSS v4.0.1 self-attested. Start at Bland's trust and security page for the Trust Portal, and see the best conversational AI platforms for a ranked comparison.
One customer result:
Idaho Housing and Finance Association saves $750,000 a year by replacing its IVR with Bland, which handles 4,000 inbound calls a day and cut average call time by 20%.
"We save thousands of dollars a day."
Greg Blake, Chief Information Officer, Idaho Housing and Finance Association
What does conversational AI on Bland cost?
Bland's pricing is $0.14 a minute on the self-serve Start plan. Enterprise pricing is custom, set with Bland's sales team. Phone is billed separately, on a customer's own carrier or Bland's at pass-through cost.
Frequently asked questions
What does open source mean in conversational AI?
Open source means the code of a layer is published under a license that lets a team read it, run it, and change it. In a conversational AI stack, a team can run an open-source framework itself and bring its own telephony, models, and monitoring, or choose a proprietary platform that runs those layers.
Is Bland open source?
No. Bland operates a closed-source model it developed itself. Bland self-hosts its models and uses no third-party model providers.
Does Bland use third-party models such as GPT or Claude?
No. Bland self-hosts its models and uses no third-party model providers.
Can I use my own phone carrier with Bland?
Yes. SIP lets you bring your own carrier, and calls from your carrier, PBX, or contact center reach a Bland agent. Phone is billed separately, on a customer's own carrier or Bland's at pass-through cost.
Do I need engineers to run a conversational AI agent on Bland?
No. Bland builds AI phone agents that answer inbound calls and place outbound calls. Your team builds and edits them in plain English with Norm or in Bland's visual builder, with no engineers needed on your side. Enterprise customers also get Bland's forward-deployed engineers.
Can engineers extend Bland with their own systems?
Yes. Anything beyond the connections in Bland's dashboard connects through Bland's REST API and webhooks, which Bland's forward-deployed engineers can help with on Enterprise. See Bland's documentation for webhook nodes that run in the middle of a call.
How was this page put together?
This page draws on LiveKit's published pages as captured on 5 and 20 November 2026, on Bland's documentation as captured on 20 November 2026, and on Bland's fact file. It describes how each approach works as those sources state it. No calls were tested or timed for it.
