A curated index of the tools developers actually use to build with AI. No models — just the tooling around them: frameworks, vector databases, inference, observability, agents and more.
Agents need access to the outside world beyond the model: scraping, search, validated outputs, inboxes, notifications and social APIs. These tools give agents reliable data and capabilities.
Crawl any website into clean, LLM-ready data.
Firecrawl scrapes single pages or entire sites, handles JavaScript rendering, and returns markdown or structured data that is ideal for RAG ingestion — no headless-browser babysitting. It turns the messy web into model-ready input.
Web search API designed for AI agents.
Tavily returns concise, relevant, LLM-friendly search results instead of raw listings, saving tokens and making agents more reliable. It has become the default search tool in many agent frameworks.
Validate and repair LLM outputs in production.
Guardrails wraps model calls with typed validators — structure, tone, safety, custom Python checks — and automatically re-asks or rephrases when a constraint fails. It keeps LLM features from shipping embarrassing outputs.
Email inboxes and cloud file storage for AI agents.
Revdoku gives every agent or project its own private bucket with a dedicated email address. Incoming messages and attachments are saved as readable files (message.md, message.json and the original .eml), so agents can read mail, collect invoices or monitor replies with the same tools they use for any other file. Buckets keep file versions, so changes can be reviewed and restored. Revdoku only receives email; it does not send.
Web scraping and structured data API for AI agents.
Context.dev turns any URL into clean Markdown or HTML, crawls whole websites, and extracts structured data that matches a JSON Schema you define. It also returns brand data such as logos, colors and fonts, so one REST API covers most of what an agent needs from the web.
One social media API for 15 platforms, built for AI agents.
Zernio replaces a dozen separate platform integrations with a single REST API for publishing, scheduling, DMs, comments, analytics and ads across Instagram, TikTok, X, LinkedIn, YouTube and more. It handles OAuth, media formatting and rate limits for you, and ships a hosted MCP server with 280+ tools so agents can run your social accounts.
A general AI agent that plans and builds the whole project.
Runable takes a plain-language goal, researches and plans it, then builds the deliverables end to end: websites, slide decks, documents, images and video. It routes work across several AI models and supports installable agent skills, so one workspace covers a lot of tasks that normally need separate tools.
Notification infrastructure for in-app, email, SMS, push and chat.
Knock lets you design cross-channel notification workflows once and deliver them everywhere through a single API. It includes a drop-in in-app feed, user preference management, batching and digests, and integrations with Slack and Microsoft Teams. It's useful when AI features need to notify users about long-running or background tasks.
LLM frameworks & SDKs →Inference & model routing →Vector databases →AI coding tools →LLM observability & evals →
Built a tool that helps developers work with AI? Send it over — we review every submission and feature the good ones on aihub.dev.
Your submission has been received. We will review it, and if it is a good fit for aihub.dev we will feature it on the site.