Agent framework connectivity
Use LangGraph Server or the OpenClaw gateway today. Provider interfaces leave room for more protocols without changing how you speak to the app.
Connect your voice to your agent framework.
Open Assistant is a voice and chat client for agentic frameworks. Connect to LangGraph over HTTP/SSE or an OpenClaw WebSocket gateway. More providers are planned.
VOICE INPUTCHAT INPUT
Open Assistantone client, multiple agent backendsLangGraphHTTP / SSEOpenClawWebSocket gateway
Architecture diagram, not a product screenshot
One client. A stack
you can shape.
Connect an agent, choose your voice services, and put its output where you need it.
Modular voice stack
Wake detection, speech recognition, and text-to-speech are selected independently. Use system speech or a supported external voice service.
Multi-device meshing
Connect devices to the same assistant experience across personal and shared screens.
Hub dashboards
A provider-backed dashboard can fill a wall-mounted tablet with relevant information and controls.
Widgets
Provider-authored resources support lists, snapshots, and actions. A refresh reads the resource without starting a model run.
Hands-free voice
Wake detection starts a continuous conversation. You can also dictate a message and review it before sending.
Interface examples are illustrative, not app screenshots. Availability of provider-backed features depends on the connected server.
Provider compatibility
Core conversation support is not the same as support for every provider resource. Check your server's capabilities before enabling an extension.
| Capability | LangGraph | OpenClaw | Future providers |
|---|---|---|---|
| Conversation | Supported: Threads, history, suggestions, speaker claims | Supported: Threads, historyNot supported: Suggestions, speaker claims | Planned: Threads, history, suggestions, speaker claims |
| Streaming & runs | Supported: Concurrent streamed runs, stream modes, background runs, client-tool pauses | Not supported: Concurrent streamed runs, stream modes, background runs, client-tool pauses | Planned: Concurrent streamed runs, stream modes, background runs, client-tool pauses |
| Dynamic UI | Supported: Dynamic UI, dedicated agents, shortcuts | Not supported: Dynamic UI, dedicated agents, shortcuts | Planned: Dynamic UI, dedicated agents, shortcuts |
| Media | Supported: Photos, images, live frames | Not supported: Photos, images, live frames | Planned: Photos, images, live frames |
| Provider resources | Supported: Widgets, dashboards, routines, connectors, tasks | Not supported: Widgets, dashboards, routines, connectors, tasks | Planned: Widgets, dashboards, routines, connectors, tasks |
| Feedback | Supported: Feedback | Not supported: Feedback | Planned: Feedback |
✅ Supported in the client · ❌ Not supported by this integration · 🔜 Provider support planned, not committed. LangGraph resource features require a server that implements the relevant provider routes.
Bring your agent
into the room.
Read the source, connect a provider, and make the voice stack yours.
View the Flutter project ↗