Voice agent pricing and keys
An agent in a meeting always runs on a MeetStream bot. What you pay on top of the bot depends on who runs the agent. The setups follow the three ways to integrate: your own stack first, then the managed agent (API name: MIA).
The bot price is 0.25 per hour at volume). If the bot also makes a post-call transcript with the MeetStream engine, add 0.10 is waived.
In a managed agent you can mix setups 3a and 3b: each pipeline section (transcriber, model, voice) uses either an inbuilt key or your own key.
2a. Bridge server
Use a bridge server when you want to run every model yourself, hosted or local (for example Parakeet for speech-to-text, a local LLM and a local text-to-speech model). The bot opens two WebSockets to your server: live_audio_required.websocket_url for the live audio and socket_connection_url.websocket_url for the control socket. You do not create a MIA agent.
Price. Bot price only. MeetStream charges no agent infrastructure and no model tokens.
See Bridge Server Architecture and the open-source starter meetstream-ai/meetstream-agent.
2b. Your own agent
Use your own agent when you already run an agent (Claude Code, Codex, Hermes, OpenClaw, GitHub Copilot CLI and others) on your own subscription, and you want it to be the brain of the voice agent. MIA still joins the call, listens and speaks. Your agent writes the answers. Set model.provider to "local_agent" and connect your agent with the open-source MeetStream Agent Kit.
Price. Bot price + $0.25 per hour for the agent infrastructure. MeetStream charges no model tokens for your agent’s reasoning: your agent uses your own subscription. Speech-to-text and voice work as for any other agent: with your own keys your provider bills you, and with inbuilt keys you pay the per-hour price of those models.
Use this setup with care. What people say and type in the meeting reaches an agent that has access to your computer and your terminal. Someone in the meeting can try to give your agent instructions (prompt injection). Keep the access level as low as possible: readonly is the default. Keep the activation code on. Your computer, your agent and the bridge must stay on for the whole call.
model.model and model.system_prompt are optional for local_agent. transcriber and voice are required. For the bridge setup, the pairing token, the activation code, the access levels and the connection test, see MIA Screen Sharing and Custom Agents.
3a. Managed agent: MeetStream’s inbuilt keys
Use MeetStream’s inbuilt keys when you do not want to open accounts with model providers. Today the inbuilt keys cover:
- OpenAI: the LLM, and OpenAI speech-to-text and voice models.
- Deepgram: speech-to-text.
For any other provider, add your own key (setup 3b).
Price. Bot price + $0.25 per hour for the agent infrastructure + the per-hour price of each model that you pick. If you pick only the LLM with an inbuilt key, you pay only for that model. The agent editor in the dashboard shows the per-hour price of your choices before the bot runs. MeetStream charges that rate for the time the bot is in the meeting.
Request shape
Set "key_source": "meetstream" on each pipeline section that uses an inbuilt key. Do not send api_key in that section.
Create an agent that uses inbuilt keys for all three sections:
Then send a bot with the agent_config_id from the response:
The models that you can use with inbuilt keys can change. If you pick a model that has no inbuilt-key price, the API returns 400 with the models that you can use (see Errors). The dashboard agent editor shows the same list.
When inbuilt keys are not available on your account
Inbuilt keys may not be available on your account yet. Then POST /api/v1/mia and PUT /api/v1/mia return 400:
The message names the section: transcriber, model or voice. If you send a bot with a saved agent that uses inbuilt keys, create_bot returns 400 with a message that names the sections, for example MeetStream-managed keys are turned off. Edit the agent to use your own keys for: transcriber, model. To fix it:
- Add your provider key in the dashboard under Configure > Integrations (tabs Models, Voice and Transcription).
- In that section, remove
key_source(or set it to"user"). - Send the request again. Do not retry the same request in a loop: it gets the same answer.
3b. Managed agent: bring your own keys
Use your own keys when you want a provider or a model that the inbuilt keys do not cover, or when you want the provider to bill you directly. You can pick any provider in the MIA API Guide, for example OpenAI, Anthropic, Google, xAI, Deepgram, AssemblyAI and ElevenLabs.
Price. Bot price + $0.25 per hour for the agent infrastructure. Your provider bills you for model usage on your own account.
- Add each provider key in the dashboard under Configure > Integrations (tabs Models, Voice and Transcription). The agent configuration holds provider names and settings, never the keys.
- Create the agent without
key_source(or with"key_source": "user"). This works in pipeline mode and in realtime mode.
If a provider key is missing, the API returns 400 API key not configured for: .... Add the key under Integrations, or use an inbuilt key for that section.
Errors
These 400 errors are about keys. The body is { "message": "..." }.
To check your prepaid balance and add credit, go to app.meetstream.ai/billing. The read-only GET /api/v1/billing/balance endpoint is rolling out: until it is enabled for your account, it returns 404.
Related
- What is MIA?: modes, providers and features.
- Create an Agent: set up an agent in the dashboard.
- MIA API Guide: every field of the agent configuration.
- MIA Screen Sharing and Custom Agents: screen tools, the bot desktop and your own agent.
- Three ways to integrate: recording only, your own stack, or a managed agent.
