Agents

Autonomous research agent that reads papers, runs experiments and compiles a LaTeX report with no human in the loop

Open-source AI agent for building apps and sites, doing research, and producing documents and slides, runnable locally with your own keys.

Local-first macOS desktop app that runs one task across several agent CLIs, collects their findings and gates file writes behind approval

Open-source browser agent that drives a real browser through a page to finish a task and returns the answer

Personal agent on your own machine with file-based memory, a background consolidation engine and skills that are tested before use.

Microsoft AI Frontiers experimental agent for browser and local file tasks, built to run on small models.

Open-source agent harness from MindsHub that runs as the default agent in its Cowork app and standalone in a terminal.

General-purpose AI agent app for Android, iOS and macOS, built on the Pi framework and extended with Pi extensions.

Personal desktop AI agent for files, writing, planning and daily tasks, with companions, plugins, scheduled runs and MCP.

Open-source, local-first desktop workspace for running Claude Agent SDK, DeepSeek Harness or Codex agents with scheduled tasks.

Self-hosted AI investment research assistant for A-shares, Hong Kong and US stocks, with data and chats kept on your own machine.

AI agent that holds a USDC wallet, pays for each model call and paid API, and runs tasks within a budget.

Iterative search agent that keeps searching, reading web pages and reasoning until it finds an answer or runs out of budget.

Autonomous research agent that forms hypotheses, edits code and runs experiments, keeping only gains that hold on held-out data

Brief-driven runner that keeps coding agents working across sessions and asks only for decisions that are yours.

Open-source Devin alternative that runs Claude Code, Codex, OpenCode or Pi in isolated workspaces you can self-host.

Open-source agentic language model (30.5B parameters, 3.3B active per token) trained for long, multi-step web information seeking.

Browser workbench that coordinates Claude Code, Codex, Gemini, OpenCode and other CLI agents as a visible team of local processes.

Sakana AI framework for automated research where LLMs generate ideas, run experiments and write papers.

Python command-line pairing agent for ML engineers that builds baselines, works Kaggle-style tasks and writes weekly reports.

Multi-agent research workbench that searches arXiv, OpenAlex and Semantic Scholar and drafts evidence-grounded literature reviews.

Open-source end-to-end general multi-agent product that answers queries and produces reports, slides and code.

Desktop AI agent that runs on your machine, edits project files and runs commands, and produces Word, Excel, PowerPoint and PDF files.

Open-source deep research agent for multi-step web research, with models published on Hugging Face.