Awesome AI Agents › IDE Integrations
Codex Autoresearch is a plugin for the Codex coding agent that turns vague improvement requests into measured experiments. The user states a goal, a benchmark command, a primary metric, correctness checks, an edit scope and a budget; the plugin measures a baseline before any code changes, then runs experiments one at a time and records what each attempt did to the metric. The benchmark must print at least one line of the form METRIC seconds=12.34, and secondary metrics can be tracked to catch trade-offs such as lower runtime at the cost of much higher memory use. The loop is setup, doctor, next, log, state, finalize-preview: setup records the configuration, doctor verifies the benchmark is trustworthy before the first experiment, next runs a single benchmark packet, log classifies the outcome as baseline, keep, discard or failure, state reads the ledger and proposes what to do next, and finalize-preview shows read-only which work could become review branches. Durable session records are stored in the target project; inside a Git repository transient packet state lives under .git/autoresearch/, otherwise it falls back to local worktree files. Git side effects are constrained: a plain measurement never stages, commits or reverts anything, keeping a result can create a commit limited to configured paths, and discards or failed checks clean only the configured experiment paths. Review branches are created only after explicit approval. An optional read-only dashboard shows metric history, the current blocker, the next action and finalization state, while the CLI remains the only writer. A quality-gap mode replaces the numeric metric with a checklist of source-backed findings for docs, UX or architecture work. It is installed through the Codex plugin picker or the TheGreenCedar/AgentPluginMarketplace source, and is inspired by karpathy/autoresearch and pi-autoresearch.
https://github.com/TheGreenCedar/codex-autoresearch
Continue is an open-source platform that enables developers to create, share, and use custom AI code assistants within popular IDEs, enhancing coding productivity and collaboration through features like Chat, Autocomplete, Edit, and Agent.
Coder is a platform for creating secure, self-hosted cloud development environments that streamline developer onboarding and optimize resource usage through automation and integrations.
Sweep is an AI-powered coding assistant plugin designed to enhance productivity and coding efficiency within JetBrains IDEs.
Claude Code plugin that distills reusable skills from real coding sessions, merges same-scenario skills instead of duplicating them, and prunes skills that stop being used.
LMQL is a Python-superset programming language designed for constraint-guided and efficient interaction with large language models, enabling advanced control flow, decoding, and integration with multiple LLM providers.
Aide is a Visual Studio Code extension that leverages AI to provide one-click code conversion, commenting, optimization, and batch processing features to enhance developer productivity.
Plugin marketplace and shared configuration for Claude Code, Codex CLI, Cursor and Gemini CLI, bundling skills, slash commands, hooks, subagents and MCP servers.
VS Code extension that runs twelve AI coding agents in one chat interface, with a brainstorm mode where two agents collaborate, selectable personas, permission controls and at-mention routing.