Agent Infrastructure

Factory provides Droid, an AI agent you can use in coding workflows and IDE integrations like Zed. It connects through login or API key setup and charges usage with Standard Tokens across multiple models.

Runloop AI provides sandboxed devboxes for agent workflows, including turn-based interaction through GitHub pull requests. It’s aimed at developers building coding agents that need to execute commands, keep state across turns, and respond to reviewer comments.

Open Interpreter is a desktop AI agent that helps you work with code, documents, and files instead of just chatting about them. It can also be run in sandboxed environments like Docker or E2B for safer execution.

Decagon is a customer support AI platform with voice and guardrails built for enterprise use. It helps teams deploy agents across support workflows with authentication, encryption, audit logs, and resiliency controls.

Fay can send real-time webhook events when research status changes, completes, or fails. It is useful if you want to trigger downstream workflows from research results instead of polling for updates.

LocalAI’s Realtime API lets you build voice and text experiences over WebSocket or WebRTC using an OpenAI-compatible protocol. It is aimed at developers who want a self-hosted, configurable realtime layer with their own VAD, STT, LLM, and TTS components.

Kiro autonomous agent stores task context, chat, and code changes to carry out multi-step work. It is hosted on AWS and designed for developers who want an agent that can retain context while it works.

Relevance AI is a low/no-code platform for creating AI agents that can complete tasks on your behalf. It’s aimed at teams that want to automate support, sales, and internal workflows without building everything from scratch.

Adds long-term memory to conversational apps built with the Vercel AI SDK. Use it to store and retrieve user context across chats and keep responses consistent over time.

Anyscale is a fully managed Ray platform that removes the infrastructure work from building and deploying AI applications. It helps teams run Ray jobs, services, and workflows with autoscaling, monitoring, and API-driven cluster management.

Open WebUI is a self-hosted AI platform for running and organizing chat across local and cloud models. It also lets you extend workflows with Python and share prompts, tools, and functions through its community.

Composio’s schema modifiers let you rewrite a tool’s schema before it reaches an agent. Use them to adjust descriptions, add or hide parameters, and set defaults when the raw tool schema needs guardrails.

BuyWhere gives developers a normalized product catalog API for Singapore and Southeast Asia. It helps AI agents search, compare, and route commerce queries without scraping storefronts.

LM Link lets you access models running on other devices as if they were local. It is built for LM Studio users who want to load remote models through the app, local server, API, or SDKs without exposing devices to the public internet.

Ollama helps you install, run, and launch open models from the terminal, with integrations for coding, chat, RAG, and automation tools. It emphasizes keeping data on your own machine while still offering cloud hardware for larger models.

Connect a Lovable app to Twitch so it can read live stream data, track channel activity, and send chat messages. It’s useful for overlays, dashboards, and creator tools that need Twitch API access.

GroqCloud is an AI inference platform for developers that focuses on low latency and predictable spend. It provides API access to text, audio, vision, and image-to-text models, with free, developer, and enterprise plans.

Fireworks AI is a model hosting and inference platform for teams building with open and proprietary models. It covers serverless inference, fine-tuning, embeddings, speech-to-text, and on-demand GPU deployments.

Together AI is a cloud platform for running, fine-tuning, and deploying open-source AI models. It is aimed at developers and teams that need model inference, GPU compute, storage, and training infrastructure in one place.

llamafile packages an LLM and runtime into one file you can download and run locally. It is aimed at developers and end users who want offline, no-install model execution across common operating systems.

Replicate lets you run and fine-tune models, and deploy custom models through an API. It’s aimed at developers who want to add image, speech, music, video, or LLM capabilities without managing model hosting themselves.