Collection · Seven decisions, one afternoon

A first agentic setup

7 picks · 7 of the atlas's 14 topics · 868,499 combined stars · snapshot 2026-10-22

The atlas has 1,294 projects in it, which is the wrong number to start from. This is one working setup instead: an agent, the things it loads, the place it runs, and the two tools that tell you afterwards whether it did what you asked. Every slot is filled once. Swap any of them later -- the point of the set is that you can start today and still know what each piece is for.

Open all 7 in the atlas →

From there: save them to your own projects, or export the set as Markdown, a standalone HTML page, or a PDF.

Runs on: Windows WSL2 macOS Linux Docker

  1. 1 The agent

    OpenCode

    anomalyco/opencode

    Terminal-native, MIT, and the one coding agent here that three separate source lists picked independently. It runs against whichever model you already pay for rather than tying the setup to one vendor.

    Open-source terminal-native AI coding agent with 131K+ stars and 2.5M+ monthly active developers. Provider-agnostic architecture supports 75+ LLM providers plus native LSP auto-configuration, multi-session parallel agents, and MCP extensibility. The build/plan agent split and client/server architecture make it the most complete open-source reference for a terminal-first coding harness

    npm i -g @opencode/cli@latest # or bun/pnpm/yarn

    219,674 stars · 4 of 39 lists · TypeScript · MIT · pushed 2026-10-22

    Windows WSL2 macOS Linux Docker

    Coding Agents opencodeMCP

  2. 2 What it knows how to do

    Superpowers

    obra/superpowers

    The most-starred skills bundle in the atlas and, unusually, one that spans seven different harnesses -- so the competencies survive changing your mind about the agent above.

    by Jesse Vincent - A strong bundle of core competencies for software engineering, with good coverage of a large portion of the SDLC - from planning, reviewing, testing, debugging... Well written, well organized, and adaptable. The author refers to them as "superpowers", but many of them are just consolidating engineering best practices - which sometimes does feel like a superpower when working wit

    git clone https://github.com/obra/superpowers.git

    303,157 stars · 7 of 39 lists · Shell · MIT · pushed 2026-10-21

    Windows WSL2 macOS Linux Docker

    Agent Skills Claude CodeClaude / AnthropicMCPCodex / OpenAIGemini / GoogleGitHub CopilotCursor

  3. 3 What it remembers

    mem0

    mem0ai/mem0

    A memory layer you drop in rather than a database you design. Apache-2.0, and it states support on Windows and in Docker, which most of this category does not.

    Mem0 is an intelligent memory layer for Large Language Models that enhances personalized AI experiences by retaining and utilizing contextual information across various applications. github | website | docs | discord | twitter | github profile | linkedin

    npm install -g @mem0/cli # or: pip install mem0-cli

    67,600 stars · 8 of 39 lists · Python · Apache-2.0 · pushed 2026-10-21

    Windows WSL2 macOS Linux Docker

    Frameworks & SDKs Codex / OpenAI

  4. 4 How it reaches your tools

    modelcontextprotocol/servers

    modelcontextprotocol/servers

    The reference MCP servers, from the people who wrote the protocol. Start here and you learn the shape every third-party server in the atlas is copying.

    Anthropic's official reference MCP server implementations (GitHub, Slack, Postgres, Puppeteer, etc.). The authoritative source for understanding correct MCP server structure before building your own

    npx -y @modelcontextprotocol/server-memory

    91,453 stars · 6 of 39 lists · TypeScript · no licence stated · pushed 2026-10-19

    Windows WSL2 macOS Linux Docker

    MCP Servers Claude / AnthropicMCP

  5. 5 Where it runs

    Daytona

    daytonaio/daytona

    Container sandboxes that start in under a tenth of a second, which is the difference between isolating every run and isolating the runs you remember to. An agent with shell access and no sandbox is a decision, not a default.

    OCI-container sandboxes with sub-90ms startup, built-in Git operations, LSP support, and indefinite state persistence. Complements E2B for harnesses that need long-lived working directories across multiple agent sessions rather than ephemeral code execution

    pip install daytona

    71,510 stars · 2 of 39 lists · no licence stated · pushed 2026-07-24

    Windows WSL2 macOS Linux Docker

    Sandbox, Security & Governance

  6. 6 What it actually did

    Langfuse

    langfuse/langfuse

    Self-hostable tracing, so the record of what the agent sent and got back stays on your own machine. Two source lists agree on it and it is the most widely adopted of the open options.

    The most widely adopted self-hostable LLM observability platform: traces every agent step, manages prompt versions, and runs evals in one tool. Preferred over cloud-only alternatives when data residency or cost control is a constraint

    pip install langfuse openai

    36,148 stars · 6 of 39 lists · TypeScript · no licence stated · pushed 2026-10-21

    Windows WSL2 macOS Linux Docker

    Observability & Evals Codex / OpenAILangChain / LangGraph

  7. 7 What to read while it runs

    Learn Claude Code

    shareAI-lab/learn-claude-code

    A read-through of how a real coding agent is built rather than a tutorial for using one. It is the fastest way to stop being surprised by the six choices above.

    by shareAI-lab - A really interesting analysis of how coding agents like Claude Code are designed. It attempts to break an agent down into its fundamental parts and reconstruct it with minimal code. Great learning resource. Final product is a rudimentary agent with skills, sub-agents, and a todo-list in roughly a few hundred lines of Python

    npm i -g @shareai-lab/kode

    78,957 stars · 3 of 39 lists · Python · MIT · pushed 2026-09-28

    Windows WSL2 macOS Linux Docker

    Docs, Learning & Lists Claude CodeClaude / Anthropic

The whole set, in one paste

Every install command above, in the order they are listed. Read them before you run them — they are other people's one-liners, and none of this set pipes a download into a shell.

npm i -g @opencode/cli@latest # or bun/pnpm/yarn
git clone https://github.com/obra/superpowers.git
npm install -g @mem0/cli # or: pip install mem0-cli
npx -y @modelcontextprotocol/server-memory
pip install daytona
pip install langfuse openai
npm i -g @shareai-lab/kode