@erichare/astra-mcp
v2.1.0
Published
Unofficial JEStats integration for Astra DB: live database tools, interactive views, guarded writes, vector and hybrid search, data-modeling skills, and Data API examples in five languages.
Maintainers
Readme
JEStats Astra DB Plugin · @erichare/astra-mcp
Unofficial, community-maintained Astra DB integration by JEStats, not affiliated with or endorsed by DataStax or IBM. An MCP server with live data tools, guarded writes, vector and hybrid search, interactive MCP Apps views, and about 1,660 bundled Data API examples. It also includes a one-command installer for Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop, and IBM Bob.
npx -y @erichare/astra-mcp initinit detects your agents and configures each one. Its login step reuses an existing credential source or asks for an Astra application token with hidden input, lets you pick a database, and writes connection settings to a git-ignored .env. There's no restart: the server re-reads credentials on every call. Requires Node.js 20+.
Already configured the Astra CLI? Run login --profile analytics to choose a database from a named profile, or use init --no-login followed by doctor to reuse the default configuration directly. Profile login stores only the profile/config selector and connection details, keeping the token in .astrarc. The npm package name stays unchanged for compatibility.
Commands
| | |
| --- | --- |
| astra-mcp | Serve MCP over stdio (the default when stdin isn't a terminal) |
| astra-mcp init | Configure agents: --agents a,b, --project, --dry-run, --yes, --no-login |
| astra-mcp login | Connect a database: --profile, --astrarc, --astra-env, --global, --database, --keyspace, --endpoint, --token-stdin |
| astra-mcp doctor | Check credentials, connectivity, .env hygiene, and agent setup (--json) |
| astra-mcp uninstall | Undo init |
| astra-mcp serve --read-only | Serve without the write tools |
| astra-mcp bob-bundle --out <zip> | Build the IBM Bob bundle |
As an MCP server
{ "mcpServers": { "astra-db": { "command": "npx", "args": ["-y", "@erichare/astra-mcp@2"] } } }The 18 tools:
- Connect:
connection_status,list_databases - Explore:
database_overview,describe_collection,describe_table,list_vectorize_providers - Query:
find,vector_search,count,distinct_values - Build:
code_examples - Change:
insert,update,delete,create_collection,create_table,create_index,drop
Destructive operations need the user's confirmation, through elicitation or an explicit confirm argument after the user approves. Credentials come from the environment, then the project's .env, host settings, login --global, and the Astra CLI's ~/.astrarc.
The package bundles everything into one file with no runtime dependencies. Releases after 2.0.0 are published from GitHub Actions with npm provenance.
Documentation, the Claude Code and Codex plugin, and the hosted OAuth endpoint for ChatGPT and claude.ai: github.com/jestatsio/astra-db-plugin.
Apache-2.0 for this plugin's own code. Includes documentation-derived examples vendored from sl-at-ibm/astra-toolkit-skill; see the bundled NOTICE for attribution and upstream licensing status. Community project, not an official DataStax or IBM product.
