OpenAI Realtime API
View profileAI modelsCloud
CommercialWhat you can do
- Stream microphone audio and receive spoken responses in one API.
Before choosing
- Long conversations incur usage charges, so estimate audio duration before implementation.
Tools
Find tools for your task. Compare apps, models, providers and runtimes.
180tools
OpenAI Agents SDK
View profileDevelopment & infrastructureSelf-hosted
FreeWhat you can do
- Create text, sandbox, realtime, and voice agents in Python.
Before choosing
- You need Python 3.10 or newer and must configure a supported model provider.
Open WebUI
View profileLocal & self-hosted appsSelf-hosted
Free tierWhat you can do
- Give configured local or hosted model endpoints a self-hosted browser interface.
Before choosing
- It does not supply model capacity; storage, provider configuration, updates and licence conditions remain operational work.
Open Notebook
View profileLocal & self-hosted appsSelf-hosted
FreeOpen sourceWhat you can do
- Collect sources and search them in one workspace.
Before choosing
- Self-hosting requires Docker and a model key unless you configure a local Ollama model.
Onyx
View profileLocal & self-hosted appsSelf-hosted
Free tierWhat you can do
- Search connected applications with source-cited agentic retrieval.
Before choosing
- Full Onyx needs server infrastructure for indexing, workers, and blob storage.
ONNX Runtime GenAI
View profileDevelopment & infrastructureSelf-hosted
FreeWhat you can do
- Embed compatible ONNX model generation with token-by-token controls in an application.
Before choosing
- Preview APIs, model artifacts and execution providers must match; it is not a ready-made chat app.
Ollama
View profileLocal & self-hosted appsDevelopment & infrastructureSelf-hosted
FreeOpen sourceWhat you can do
- Download and run supported models through local chat, CLI and HTTP APIs, with optional cloud features.
Before choosing
- Memory and GPU fit depend on model and context. Cloud selection changes the data path. Keep the local API on loopback unless you add remote access controls.
NVIDIA System Management Interface
View profileDevelopment & infrastructureSelf-hosted
FreeWhat you can do
- Inspect driver-visible NVIDIA GPU memory, utilization and device state before local inference.
Before choosing
- NVIDIA-only diagnostics: one sample is not peak-memory evidence and the CLI table is not a stable API.
Neo4j
View profileDevelopment & infrastructureCloud · Self-hosted
Free tierWhat you can do
- Model connected entities as nodes and relationships.
Before choosing
- Check edition licensing: Community is GPL v3, while other offerings require commercial terms.
Nano Banana Pro (Gemini 3 Pro Image)
View profileAI modelsCloud
CommercialWhat you can do
- Generate images with text, data labels, and multimodal instructions.
Before choosing
- Access and product naming are changing during Google's Imagen migration; verify the current endpoint.
n8n
View profileWeb servicesLocal & self-hosted appsCloud · Self-hosted
FreeWhat you can do
- Chain APIs, models, databases, and files into visual workflows.
Before choosing
- Choose and secure hosting deliberately; credentials and workflow data follow that deployment path.
Moshi (Kyutai)
View profileAI modelsSelf-hosted
FreeWhat you can do
- Build speech-native conversations that listen and speak simultaneously.
Before choosing
- Real-time operation requires a capable GPU and research-oriented deployment work.
Ministral 3
View profileAI modelsSelf-hosted
FreeWhat you can do
- Run small open-weight models on laptops or edge hardware.
Before choosing
- Small models suit scoped tasks; test harder reasoning tasks before relying on them.
Mistral Medium 3.5
View profileAI modelsCloud
CommercialWhat you can do
- Process text and images with a large-context multimodal model.
Before choosing
- Verify the provider, region, contract, and licence for your chosen deployment path.
MiniMax-M2
View profileAI modelsSelf-hosted
CommercialWhat you can do
- Use a model focused on coding and agentic tool chains.
Before choosing
- Check the exact model version and licence before self-hosting or commercial use.
Midjourney
View profileWeb servicesCloud
CommercialWhat you can do
- Create art-directed image drafts from prompts and references.
Before choosing
- Cloud access and commercial terms can change; verify rights and final exports for each use.
mem0
View profileDevelopment & infrastructureCloud · Self-hosted
Free tierWhat you can do
- Give assistants retained facts across separate sessions.
Before choosing
- You must provide an API key and account for model and storage usage.
Make
View profileWeb servicesCloud
Free tierWhat you can do
- Connect apps through visual, multi-step automation scenarios.
Before choosing
- Designs with many operations or active scenarios may exceed the free allowance.
LTX-Video
View profileAI modelsSelf-hosted
FreeWhat you can do
- Generate video from text or images through ComfyUI workflows.
Before choosing
- Release-specific hardware, resolutions, clip lengths, licences, and nodes require verification.
Lovable
View profileWeb servicesCloud
Free tierWhat you can do
- Describe a web application and iterate against a live preview.
Before choosing
- Generated applications need review, and iterative debugging can consume credits unexpectedly.
LocalAI
View profileLocal & self-hosted appsDevelopment & infrastructureSelf-hosted
FreeOpen sourceWhat you can do
- Self-host compatible APIs and a web interface over selected local AI backends.
Before choosing
- Every feature needs a compatible backend; remote exposure requires deliberate authentication and network controls.
LM Studio
View profileLocal & self-hosted appsDevelopment & infrastructureSelf-hosted
FreeWhat you can do
- Download, chat with and locally serve compatible models from a desktop application.
Before choosing
- Model fit, extensions, updates and selected integrations can create separate resource or network paths.
llm-d
View profileDevelopment & infrastructureSelf-hosted
FreeWhat you can do
- Route Kubernetes model-server pools with inference-aware scheduling and cache-aware routing.
Before choosing
- Requires a model engine, Kubernetes, accelerator capacity and sustained platform operations.
llm-compressor
View profileDevelopment & infrastructureSelf-hosted
FreeOpen sourceWhat you can do
- Compress Hugging Face checkpoints for vLLM deployment.
Before choosing
- Compression changes model artifacts and requires checking compatibility with your serving setup.