Tag
13 catalogue entries use this topic; each Project's license policy is labeled separately.
AI coding agent for terminal, editor, and desktop
OpenCode is a coding agent that runs where you already work rather than in a separate chat window. The same session can be driven from a terminal, an editor extension, or a desktop client, and it keeps context across all three.
Run open language models locally with one command
Ollama reduced local model hosting from an afternoon of CUDA troubleshooting to a single pull command. It handles model download, quantisation selection, GPU offloading, and memory management, and exposes an OpenAI-compatible endpoint on localhost.
Turn any website into clean, model-ready data
Firecrawl handles the unglamorous part of feeding the web to a language model: rendering JavaScript, following pagination, stripping navigation and cookie banners, and returning markdown or structured JSON that a model can actually use.
One chat interface for every model you run or connect
Open WebUI is the front end that turns a local inference server into something a household or a team can actually use. It handles accounts, conversation history, document retrieval, and model switching, and speaks to both local runtimes and hosted APIs.
Inference engine that runs LLMs on almost any hardware
llama.cpp is the layer most local model tooling is built on. It is a C and C++ inference engine with no Python runtime, aggressive quantisation support, and hand-tuned kernels for a remarkably wide range of hardware.
Autonomous agents that complete real engineering tasks
OpenHands runs agents that plan and execute multi-step engineering work: reading a repository, editing files, running tests, and iterating until a task is done. Each agent works inside a sandboxed runtime rather than on the host machine.
Polished chat interface for many model providers
LobeChat is the most designed of the open chat front ends. Plugins, an assistant marketplace, voice input and output, and vision support are all present, and the interface holds up next to commercial products.
A control layer for customer-facing AI agents
Parlant addresses the problem that stops most support agents from shipping: a model that is helpful ninety-five percent of the time is not safe to put in front of customers. It manages behaviour through explicit, context-matched guidelines rather than one long prompt.
OpenAI's assistant and the interface that set the pattern
ChatGPT is the product that made conversational assistants ordinary. Beyond the chat box it now covers file analysis, image generation, web browsing, voice, and custom assistants sharing one account.
Anthropic's assistant, strongest on long documents and code
Claude is positioned around careful reasoning over long inputs. The large context window makes whole-repository and long-document work practical in a way that shorter context assistants cannot match.
An editor rebuilt around model assisted editing
Cursor forked VS Code rather than extending it, which let it change the editing model itself. Multi-file edits, codebase-wide context, and a tab completion that predicts the next edit rather than the next token are the result.
AI completion and chat inside the editor
Copilot put model assisted completion in front of a very large number of developers first, and inline suggestion in the editor remains the interaction most people mean by AI coding.
Desktop application for running language models locally
LM Studio made local models approachable for people who do not want a terminal. Browse a catalogue, download a quantisation that fits the machine, and chat, with hardware requirements explained in plain language.