Shape agent capabilities into your workflow.

Command Code builds features, fixes bugs, writes tests, and refactors inside an interactive terminal session. Its Taste system treats accepted, rejected, and edited output as signals, gradually adapting to how you prefer code to be written. Taste profiles can also be pushed and pulled for use across a team.

Devon is an open-source pair-programming agent available through a web interface or terminal UI. It can explore a repository, edit multiple files, write tests, fix bugs, and accept corrections while it works. Installation requires Node.js, pipx, and at least one Anthropic, OpenAI, or Groq API key; support for non-Python projects is limited, while Windows and local-model modes remain immature.

Serenity.skill guides an agent through supply-chain research for themes such as AI, semiconductors, and robotics. It maps demand and industry layers, looks for bottlenecks with limited suppliers, difficult capacity expansion, or long qualification cycles, then checks filings, financial reports, customers, capacity, and risks. The output is a prioritized list of companies or fund directions with an evidence trail and follow-up checks. It is strictly a research aid: it does not execute trades, promise returns, or replace the user's final investment decision.

Browserbase MCP Server gives compatible assistants a small set of tools for starting cloud browser sessions, navigating pages, performing actions, observing interactive elements, and extracting information. The hosted HTTP endpoint is the easiest setup, while self-hosted deployments can run through NPM or Docker. These deployments still require a Browserbase API key and project ID, and choosing a model other than the default also requires that provider's API key.

The OpenAI Agents SDK organizes agent workflows around tools, guardrails, handoffs, sessions, and tracing. It also includes sandbox agents for filesystem work and realtime agents for low-latency voice interactions, with support for Node.js, Deno, and Bun. Sandbox agents are marked beta, and browser voice sessions should use a server-created ephemeral client token rather than exposing a long-lived API key.

Agent Sprite Forge turns natural-language requests into a structured 2D game-asset workflow. An agent generates raw visuals, then local processors remove backgrounds, split frames, validate results, and export PNGs, sprite sheets, GIFs, layered maps, and scene data for Godot, Unity, or other 2D projects. Standard sprite and map skills require built-in image generation. The denser video-to-sprite workflow depends on Grok Build's image-to-video tools, though exported frames can be reprocessed locally with ffmpeg and Pillow.

Toad combines a traditional shell with a terminal UI for working with coding agents such as Claude, Gemini, and Codex. Its agent catalog lets you find, install, and run compatible tools from one interface, while Agent Client Protocol support gives developers a way to add more agents.

Give an Agent Skills-compatible coding tool practical SwiftUI guidance for state management, view composition, performance, layout, animation, accessibility, charts, lists, and image handling. The skill also covers adopting Liquid Glass on iOS 26 and later, and can be installed through the skills CLI, Claude Code, Cursor, or a manual skill-directory setup.

obsidian-second-brain turns an Obsidian vault into a shared working memory for Claude Code, Codex, Gemini CLI, OpenCode, Hermes, and Pi. Its commands can save conversations, ingest URLs or files, update related notes, synthesize themes, reconcile contradictions, and search the vault. Installation is platform-specific and requires pointing the agent to the vault; some research, calendar, and background workflows also depend on external services or capabilities that are not available in every supported runtime.

grok-cli connects a terminal coding agent to xAI’s Grok API, with web and X search, sub-agents, and Telegram control while the CLI keeps running. It offers an interactive terminal UI plus a headless prompt mode. A Grok API key is required, and the community project is not affiliated with or endorsed by xAI.

Harbor is a CLI and companion app that assembles local model backends, chat interfaces, and supporting services through preconfigured Docker Compose integrations. Users choose components such as Ollama, llama.cpp, vLLM, Open WebUI, SearXNG, speech, or image tools and start them together with a single command, while Harbor centralizes configuration and logs. It is intended as a local LLM development and experimentation environment rather than a production deployment platform.

Learn how an agent gateway grows from a basic model loop into a routed, proactive, and resilient system. Ten progressive Python sections add tool use, sessions, Telegram and Feishu channels, routing, memory, cron work, durable delivery, retries, and concurrency, with matching English, Chinese, and Japanese code and documentation.

NotFair is a Claude Code marketing plugin for investigating wasted ad spend, traffic losses, creative fatigue, and SEO issues through Google Ads, Meta Ads, Search Console, and related connectors. It can perform only the account changes exposed by those connectors and also includes a local CMO portal with specialist agents. Connected ad features require OAuth, and the Meta write surface is intentionally limited.

This repository states that Crystal was deprecated in February 2026 and directs users to Nimbalyst for active downloads, documentation, and updates. Nimbalyst is a local-first workspace for human and agent collaboration, with editors for code, Markdown, spreadsheets, diagrams, and visual documents. It uses Git worktrees to isolate parallel coding sessions and is built on Electron.

AutoCodeRover tackles GitHub bug reports and feature requests by using program-structure-aware search to locate relevant files, classes, and methods before generating a patch. When a test suite is available, it can use fault localization and regression testing to improve patch selection. It works on local projects and also includes a SWE-bench experiment mode, with a Conda environment and a supported model API key required.

Laminar is an observability platform for tracing AI agent runs, defining behavioral signals, running evaluations, building dashboards, and preparing datasets. Its SDKs instrument supported AI frameworks, while alerts can notify Slack when a described failure pattern appears. Evaluations run locally or in CI, and coding agents can inspect traces through CLI or MCP access. Teams can use the managed service or self-host with Docker; self-hosted deployments collect anonymous usage telemetry unless it is explicitly disabled.

Ralph Orchestrator drives coding agents through role-based hats, events, and iterative loops until they emit LOOP_COMPLETE or reach the iteration limit. It supports backends such as Claude Code, Codex, Gemini CLI, Kiro, and OpenCode, with planning, persistent tasks and memory, plus gates for tests, lint, and type checks. Telegram can add human guidance, each MCP server is scoped to one workspace, and the web dashboard remains Alpha.

DeepCamera is an open-source platform for running pluggable AI camera skills locally, including scene analysis, object detection, person re-identification, and fall detection. Skills use a common protocol and can target different model and hardware backends across Apple Silicon, NVIDIA, Intel, AMD, and CPU systems. Its Aegis desktop companion connects camera sources, installs skills, and delivers alerts or camera conversations through Telegram, Discord, or Slack. The catalog is active and still expanding, so capability maturity may vary by skill.