Open-source AI tools worth bookmarking
Local research agents, a Perplexity-style search engine, OCR for your documents, agent workflows, MCP servers and the lists I learn from.
A month of saved AI links, sorted by what they're actually for. Most of these run locally or self-hosted, which matters when the data is a client's.
Coding and building
opencode is an open-source coding agent for the terminal. Like Cursor, but you point it at whatever model you want.
AI Website Cloner Template clones a website with one command using AI coding agents. Great for studying how a site is built, or for rebuilding a client's old site on a modern stack. Don't use it to copy someone else's design and brand.
Manus builds apps that come with a working backend, not just a front-end mockup.
CLI-Anything turns apps like Blender or OBS into command-line tools that AI agents can drive and version-control as text. If you've wanted an agent to operate desktop software, this is the missing bridge.
Research and search
AgenticSeek is a fully local, Manus-style agent. It thinks, browses the web and writes code, and the only bill is electricity.
Vane is an answer engine like Perplexity. It searches the live web, reads the pages and answers with sources. Self-host it and keep your queries private.
Documents and content
PaddleOCR turns PDFs and images into structured data your AI can use. Invoices, scanned contracts and screenshots become text and tables.
PPT Master turns documents or a topic into real, native PowerPoint decks: native shapes, charts and transitions, not images pasted onto slides. You can still edit everything.
InvokeAI is a creative engine for image generation, with a proper canvas, layers and workflows on top of open models.
Workflows and agents
Langflow lets you build agent and RAG workflows visually, then deploy them as an API or an MCP server.
Awesome MCP Servers is a directory of MCP servers that let AI agents talk to browsers, databases, APIs and more. Check here before writing your own integration.
Lists I keep coming back to
- Awesome LLM Apps: 100+ ready-to-run AI agents and RAG apps with full source. The best way to learn is to read working code.
- Public APIs: a huge list of free APIs to build side projects and demos on.
- Awesome: the original list of curated lists, on nearly every topic.