For Julia / scientific computing

DepScope for Julia

Julia package intelligence for AI coding agents working on scientific computing.

The problem

AI coding agents (Claude, Cursor, ChatGPT, Copilot) recommend Julia packages based on training data 6-12 months stale. Recent CVEs missed, deprecated libraries still suggested, package names sometimes hallucinated.

Every agent also queries JuliaRegistries/General independently — billions of redundant fetches a day, tokens burned parsing JSON the model doesn't need.

One API call — live health, vulnerabilities, alternatives

curl https://depscope.dev/api/check/julia/DataFrames

Returns a health score, list of live CVEs, deprecation flags, latest version, alternatives — all from OSV + GitHub Advisory Database, cached.

For AI agents that prefer token-efficient responses:

curl https://depscope.dev/api/prompt/julia/DataFrames

Same signal, much more compact payload — less input tokens per decision.

Integrate in one line

Claude Desktop / Cursor  Remote MCP (zero install):

{ "mcpServers": { "depscope": { "url": "https://mcp.depscope.dev/mcp" } } }

Legacy stdio  Local MCP server on npm:

npm install -g depscope-mcp

GitHub Actions CI  Audit your Julia dependencies on every PR:

- uses: cuttalo/depscope@main
  with:
    ecosystem: julia

Any language  Just call the API. No auth, 200 req/min.

Next steps

→ Try the API now with DataFrames

Package intelligence is infrastructure. DepScope is the shared layer so every AI coding agent — and every developer — can rely on the same live data. Open infrastructure, MIT, EU-hosted.