Tools
Apps, editor extensions, command line tools and MCP servers, written up by people who actually run them. What it does well, what it does badly, and what you would use instead.
LlamaIndex
A Python framework for retrieval only. It connects a model to your own documents, databases or APIs, and does not try to chain everything.
Tested on Claude Sonnet · Jun 2026
Slack MCP server
Lets the model read and post to the Slack channels it can access: search history, summarise a thread, or draft a message for you to send.
Tested on Claude Sonnet · Jun 2026
Vercel AI SDK
A TypeScript library for streaming model responses into a web UI. Hooks handle the streaming and state so you do not build them by hand.
Tested on GPT-5 · Jun 2026
LangChain
A Python and JavaScript framework for chaining prompts, tools and retrieval into a pipeline, with a large set of ready-made integrations.
Tested on GPT-5 · Jun 2026
GitHub MCP server
Issues, pull requests and code search without leaving the conversation. Scope the token to what you need.
Tested on Claude Sonnet · May 2026
Instructor
A small Python library that forces a model output into a Pydantic schema and retries automatically when the response fails validation.
Tested on Claude Sonnet · Jun 2026
Haystack
An open-source Python framework from deepset for production retrieval and search pipelines, with more explicit pipeline control than most…
Tested on GPT-5 · Jun 2026
Replicate
Run thousands of community-published models, image, audio, video and language, behind one API without hosting any of them yourself.
Tested on GPT-5 · Jun 2026
ChatGPT desktop app
The native macOS and Windows app: a global hotkey, screen access on request, and a persistent window in place of a browser tab you keep losing.
Tested on GPT-5 · Jun 2026
Comet
The Perplexity browser with an AI assistant built into the browser itself. It reads the open tab, summarises it and acts on the page, and does not…
Tested on GPT-5 · Jun 2026
Claude Skills and connectors
How Claude extends a conversation with packaged capabilities and live connections to services such as Google Drive and Slack. Configure them once and…
Tested on Claude Sonnet · Jun 2026
Aider
A terminal pair programmer that edits files straight in your working tree and commits each change with a generated message. No copy-paste loop.
Tested on Claude Sonnet · Jun 2026
Windsurf
A standalone code editor forked from VS Code and built around an agent. The assistant plans and runs multi-file changes and does more than suggest…
Tested on Claude Sonnet · Jun 2026
Custom GPTs
The ChatGPT way to package instructions, uploaded files and actions into a named assistant that you or your team open directly, with no API…
Tested on GPT-5 · Jun 2026
Groq Cloud
Hosted inference on custom LPU hardware, built for speed. Open-weight models answer at a token rate that makes streaming feel close to instant.
Tested on Llama 3.3 70B · Jun 2026
OpenRouter
One API and one bill for dozens of hosted models from different providers. Use it to compare models or to route each task to whichever is cheapest or…
Tested on GPT-5 · Jun 2026
Raycast AI
AI commands inside the Raycast launcher on macOS. Rewrite a selection or answer a question without switching apps.
Tested on GPT-5 · Jun 2026
Notion MCP server
Gives the model access to your Notion pages and databases: read project docs, update a status field, or draft a new page from a conversation.
Tested on Claude Sonnet · Jun 2026
Together AI
Hosted inference for open-weight models such as Llama, Mixtral and Qwen, plus fine-tuning, at a lower price than most proprietary-model APIs.
Tested on Llama 3.3 70B · Jun 2026
Filesystem MCP server
Gives the model read and write access to directories you name. The most useful MCP server and the one that deserves the most thought before you…
Tested on Claude Sonnet · May 2026
AnythingLLM
A desktop app for chatting with your own documents against a local or hosted model. The retrieval pipeline is built in, so you do not assemble it…
Tested on Llama 3.3 70B · Jun 2026
GitHub Copilot
Inline code suggestions and a chat panel in the editor, backed by a choice of models. Most developers try it first, for better or worse.
Tested on GPT-5 · Jun 2026
Fetch MCP server
Lets the model retrieve a URL and read it as text. Simple and useful, and the plainest case of untrusted input in the whole protocol.
Tested on Claude Sonnet · May 2026
PostgreSQL MCP server
Lets the model query your database directly. Read-only is the setting that counts, and it is not the default everywhere.
Tested on Claude Sonnet · May 2026