Skip to content
Get packs

Stack for developers

The AI-native developer stack

For engineers who want agents to do real work — not demos. Pick an open coding agent, give it battle-tested engineering skills and a spec-first workflow, connect up-to-date docs and code intelligence through MCP, and keep the boring utilities one click away.

Open source
17
Agent skills
5
Browser tools
3
Total pieces
30
Download stack kit

ZIP: setup README, install helper, the skill folders (with licences) and free prompts.

How it connects

  1. 01

    Spec first

    Write the feature spec with Spec Kit or OpenSpec; the agent plans against it instead of guessing.

  2. 02

    Build with skills

    Superpowers / engineering skills make the agent write tests first, work in small steps and self-review.

  3. 03

    Ground it

    Context7 keeps docs current; the GitHub MCP server and a code graph give it your real codebase and issues.

  4. 04

    Verify and ship

    Run the security-audit skill, test the API in Hoppscotch, review with the checklist, deploy.

1 · Coding agent & editor

Open-source agents and editors you can run locally or point at any model.

OpenCodeOpen sourceOpen-source terminal coding agent that works with many model providers.★ 224k · MITOpenHandsOpen sourceAutonomous AI developer that can plan, edit, run and test in a sandbox.★ 92.5k · MITVS CodeOpen sourceThe editor most agent extensions target.★ 194k · MITOllamaOpen sourceRun open models locally for private or offline work.★ 186k · MITClaude CodeAI toolAnthropic's agentic coding tool — first-class support for skills.PaidCursorAI toolAI-first editor if you prefer an IDE over the terminal.FreemiumCodeRabbitAI toolAI code review on every pull request.Freemium

2 · Engineering skills & methodology

Skills that teach the agent how good engineers work: plan, test, review, ship.

SuperpowersOpen sourceAgentic skills framework: brainstorm → plan → TDD → review, enforced as skills.★ 311k · MITSkillsOpen sourcePractical engineering skills straight from a senior engineer's agent setup.★ 310k · MITAgent SkillsOpen sourceProduction-grade engineering skills for coding agents.★ 113k · MITSpec KitOpen sourceSpec-driven development toolkit: write the spec, let the agent build to it.★ 145k · MITOpenSpecOpen sourceLightweight spec-driven workflow for AI coding assistants.★ 73.7k · MITsecurity-auditOpen sourceMulti-phase security audits with verified findings.★ 56.8k · MITWriting PlansAgent skillTurns a spec into a step-by-step plan before any code is written.Agent skill · MITTest-Driven DevelopmentAgent skillForces red → green → refactor on every change.Agent skill · MITSystematic DebuggingAgent skillRoot-cause debugging instead of guess-and-patch.Agent skill · MITCode Review and QualityAgent skillReviews changes like a senior engineer before you merge.Agent skill · MITAgent Browser ControlAgent skillLets the agent drive a real browser to test what it built.Agent skill · Apache-2.0

3 · Context, memory & docs (MCP)

Give the agent the right context: current docs, your codebase graph and your GitHub.

Context7Open sourceUp-to-date library docs for agents, so they stop using outdated APIs.★ 64.1k · MITGitHub MCP ServerOpen sourceOfficial GitHub MCP server: issues, PRs and code from inside the agent.★ 34.2k · MITCodeGraphOpen sourcePre-indexed code knowledge graph that stays in sync with your repo.★ 76.2k · MITAwesome MCP ServersOpen sourceDirectory of MCP servers to connect databases, browsers and SaaS tools.★ 97.2k · MITRRepository onboarding skillPromptSkill that makes an agent onboard itself to a new repository.In a pack

4 · Ship & everyday utilities

Frameworks agents are fluent in, plus the small tools you reach for daily.

Next.jsOpen sourceFull-stack React framework with first-class agent and deployment support.★ 143k · MITFastAPIOpen sourceFast, typed Python APIs that agents generate reliably.★ 105k · MITHoppscotchOpen sourceOpen-source API client for testing what the agent built.★ 80.7k · MITJJSON Formatter & BeautifierBrowser toolInspect API responses with errors pinpointed by line.RRegex TesterBrowser toolTest regexes safely with timeouts.JJWT DecoderBrowser toolDecode tokens locally — nothing leaves the browser.CCode review checklistPromptA review prompt that catches what agents commonly miss.Free prompt

Setup checklist

  • Install a coding agent (OpenCode, Claude Code or similar) and connect a model
  • Install the engineering skills (`install.sh` in the kit copies them to ~/.claude/skills)
  • Add Context7 and the GitHub MCP server to your agent's MCP config
  • Initialise Spec Kit or OpenSpec in your repo
  • Keep a CLAUDE.md / AGENTS.md with your project's conventions