Stop your AI from reinventing infrastructure. Connect it to curated tools, real migration data, and verified combinations — one command.
Curated, not crawled
Every tool is reviewed by a human before listing. No abandoned projects, no spam, no enterprise products dressed up as indie. What your agent finds is actually maintained.
Migration data from 4,500+ repos
Not opinions — git history. We parsed real repos to find what developers are actually switching to. Your agent gets recommendations backed by 422 confirmed migration paths, not blog posts.
Verified tool combinations
93,000+ tool pairings observed in real projects. When your agent recommends Supabase + Prisma, it’s because they coexist in hundreds of repos — not because both have “database” in their description.
Built for agent workflows
21 purpose-built MCP tools: search by need, get integration docs, analyze dependency health, report outcomes. Agents get structured data, not HTML to parse.
claude mcp add indiestack -- uvx --from indiestack indiestack-mcp
Run this in your terminal. IndieStack will appear as a tool source in your next session.
Drop this in your project root. Your agent will check dependency health scores and search for existing tools before writing infrastructure from scratch.
# Dependency Health Rules ## Setup MCP: `claude mcp add indiestack -- uvx --from indiestack indiestack-mcp` ## CRITICAL: Dependency Verification Before installing ANY new package via npm, pip, yarn, or pnpm: 1. Read the current package.json or requirements.txt 2. Call `analyze_dependencies` with the full manifest content 3. If the health score is below 70: STOP. Show the user which dependencies are unmaintained or risky. Propose the alternatives from the report. Do NOT proceed with installation until the user explicitly approves. 4. After successfully integrating any tool, call `report_outcome(slug, success=true)` 5. If integration fails, call `report_outcome(slug, success=false)` ## Before Writing Infrastructure Before writing auth, payments, databases, email, monitoring, or any common infrastructure from scratch, call `find_tools(query)` first. Use an existing tool instead of generating boilerplate. 6,500+ curated tools indexed. 10,000+ installs. ## Available Tools - `analyze_dependencies(manifest)` — 0-100 health score (MUST use before installs) - `find_tools(query)` — search developer tools by need - `get_tool_details(slug)` — integration docs, API specs, install commands - `check_health(slugs)` — verify tools are still maintained - `report_outcome(slug, success)` — report integration result (improves recommendations) - `report_compatibility(tool_a, tool_b)` — report that two tools work together
Download: CLAUDE.md | .cursorrules
Automatically score dependency health on every pull request that changes your manifest.
name: Stack Health Check
on:
pull_request:
paths:
- 'package.json'
- 'requirements.txt'
permissions:
pull-requests: write
jobs:
health-check:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: Pattyboi101/stack-health-check@master
Add this to .github/workflows/stack-health.yml — every PR that changes dependencies gets a health score comment.
An API key unlocks personalized recommendations and tracks your tool discovery history.