AI Tools With an API

Machine-learning fraud detection built into Stripe payments.

Stripe Radar uses machine-learning models trained across millions of businesses on Stripe to score each payment for fraud risk in real time, automatically blocking or flagging suspicious transactions without merchants having to build their own fraud rules.

• Real-time ML fraud scoring per transaction
• Trained across Stripe's entire network
• Customizable rules on top of ML risk scores

Fast inference API for running open-source AI image, video, and audio models.

fal.ai provides an API for running popular open-source generative models (image, video, and audio) with very low latency, letting developers add AI generation features to their apps by calling an endpoint instead of hosting and optimizing models themselves.

• Low-latency inference for popular open models
• Simple pay-per-use API pricing
• Supports image, video, and audio generation models

Framework and platform for orchestrating teams of collaborative AI agents that complete multi-step tasks together.

CrewAI lets developers define multiple AI agents with specific roles, tools, and goals, then coordinates them as a "crew" that collaborates to complete complex workflows. It offers both an open-source Python framework and a hosted platform for deploying, monitoring, and scaling multi-agent automations in production.

• Multi-agent role-based orchestration
• Open-source core framework
• Hosted deployment and monitoring platform
• Tool and API integrations per agent

Evaluation and observability platform for building reliable LLM applications.

Braintrust gives teams a way to systematically test and score LLM application outputs against datasets, track regressions across prompt/model changes, and monitor production logs — aimed at catching quality issues before they reach users.

• Systematic LLM output evaluation and scoring
• Regression tracking across prompt/model changes
• Production log monitoring

On-demand GPU cloud for AI training, fine-tuning, and inference.

RunPod provides affordable, on-demand GPU instances for AI workloads — training, fine-tuning, and hosting model inference endpoints — aimed at developers and small teams who need serious GPU power without long-term cloud contracts.

• Wide range of GPU types at competitive pricing
• Serverless inference endpoints
• Community and secure cloud tiers

AI speech recognition API supporting many languages and accents.

Speechmatics provides a speech-to-text API built on self-supervised AI, aimed at high accuracy across accents, dialects, and noisy audio for developers building transcription features.

• Multi-language, multi-accent transcription
• Real-time and batch transcription API
• High accuracy in noisy audio

Open-source LLM observability platform for monitoring cost, latency, and usage.

Helicone is a lightweight, open-source observability layer that sits between an application and its LLM provider, logging every request to track cost, latency, and errors, and giving teams dashboards and alerts without changing much application code.

• One-line integration via proxy or SDK
• Open-source and self-hostable
• Cost, latency, and usage dashboards

Framework from LangChain for building stateful, multi-step AI agents.

LangGraph is a library from the LangChain team for building AI agents as explicit graphs of steps rather than a single prompt-response loop, giving developers control over branching logic, retries, and long-running multi-agent workflows.

• Graph-based agent orchestration
• Built-in state persistence and retries
• Works with any LLM provider

AI-powered web data extraction platform that builds a structured knowledge graph.

Diffbot uses computer vision and NLP to automatically extract structured data (articles, products, discussions) from any webpage without custom scraping rules, and aggregates it into a queryable knowledge graph of facts about people, organizations, and products.

• Automatic structured extraction, no scraping rules needed
• Queryable web-scale knowledge graph
• Computer vision + NLP-based page understanding

Low-latency voice AI infrastructure for building real-time speech and voice-agent products.

Cartesia provides real-time text-to-speech and voice models built for low latency, designed for developers building voice agents, live dubbing, and interactive audio products. Its API-first approach targets applications where response speed and voice quality both matter, such as live customer support calls.

• Very low-latency text-to-speech
• Real-time voice cloning options
• Built for voice-agent products
• Developer-first API

AI lip-sync tool for dubbing video into other languages realistically.

Sync Labs re-syncs a speaker's lip movements to match dubbed audio in another language, aimed at making translated videos look natural rather than obviously dubbed.

• AI lip-sync for dubbed video
• API for developer integration
• Supports many source/target languages

Web data infrastructure platform providing clean, structured data for AI models.

Bright Data provides proxy networks and web-scraping infrastructure at scale, increasingly marketed as an AI data pipeline — turning public web content into clean, structured datasets for training and grounding LLMs and AI agents.

• Large-scale proxy and scraping infrastructure
• AI-ready structured data output
• Pre-built datasets for common data types

AI-powered search and discovery API for websites and apps.

Algolia provides a fast, typo-tolerant search API that developers drop into e-commerce and content sites, now extended with AI features like natural-language query understanding, personalized ranking, and generative answers over a site's own content.

• Millisecond-fast, typo-tolerant search
• AI-powered relevance and personalization
• Generative answers over site content

Open-source Python library for building and sharing AI model demos.

Gradio (built by Hugging Face) lets developers wrap a machine learning model in a simple web UI with a few lines of Python — used across the AI community to demo models, collect feedback, and share interactive AI apps without frontend work.

• Build a model UI in a few lines of Python
• One-click shareable public links
• Huge library of AI demos built with it

Autonomous AI software engineer that plans, codes, tests, and ships pull requests on its own.

Devin, built by Cognition Labs, is an autonomous coding agent that works in a sandboxed environment with its own shell, browser, and code editor. Given a ticket-style task, it plans the approach, writes code across multiple files, runs the test suite, debugs failures, and opens a pull request for human review. It supports running many tickets in parallel, letting engineering teams delegate well-scoped backlog items instead of implementing them by hand.

• Autonomous end-to-end coding sessions
• Opens PRs with tests included
• Runs many tickets in parallel sandboxes
• GitHub and Slack integration

Emotionally intelligent voice AI platform that reads and responds to vocal tone and expression.

Hume AI builds voice and language models trained to detect emotional expression in speech and respond with emotionally appropriate tone. Its Empathic Voice Interface (EVI) API lets developers add conversational voice agents that adapt their delivery based on how a user sounds, aimed at more natural-feeling voice interactions.

• Emotion-aware voice responses
• Empathic Voice Interface (EVI) API
• Real-time conversational voice AI
• Expression measurement research tools

Web search API designed for AI agents and RAG applications.

Tavily is a search API built specifically for LLM agents to answer questions — it returns pre-summarized, source-cited results optimized for feeding directly into a model's context window, rather than raw search-engine results pages meant for humans.

• LLM-optimized, pre-summarized search results
• Source citations included in results
• Built for real-time agent research tasks

Data framework for connecting LLMs to your own documents and data.

LlamaIndex is an open-source data framework that helps developers load, index, and query private or domain-specific data with large language models — the connective layer commonly used to build retrieval-augmented-generation (RAG) pipelines and AI knowledge agents.

• Data connectors for 100+ sources
• RAG pipeline building blocks
• Agent and workflow orchestration

High-speed AI image and video generation API built on custom inference infrastructure.

Runware provides an API for running popular open image and video generation models (Stable Diffusion, FLUX, and others) on custom-built inference infrastructure optimized for very low latency, aimed at developers who need fast generation at scale without managing GPUs themselves.

• Very low-latency image/video generation
• Supports many popular open models
• No GPU infrastructure to manage

Real estate data platform that unifies property data with AI-powered insights.

Cherre connects and cleans property, ownership, and market data from dozens of public and private sources into a single graph, then applies AI to surface deal opportunities, valuation insights, and portfolio risk for real estate investors and lenders.

• Unified real estate data graph
• AI-driven deal and valuation insights
• Portfolio-level risk analytics