M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
S

SpotDB MCP Server

Community
Give your agents a secure, containerized SQL sandbox
How to install ↓ Compare GitHub source repository ↗
Category
Database #50 of 58
Stars
★ 21 Popular
Transport
stdio (local process)
Runtime
Prebuilt binary · Node.js
Credentials
Optional API key
License
MIT
Last commit
⚠
45FMRS · D

SpotDB provides a lightweight, secure sandbox with both MCP and API access, suitable for AI and data exploration, but note data ephemerality.

Strongest · Setup experience 10/20 Weakest · Security and permissions 8/20

Reliability
9/20
Security and permissions
8/20
Maintenance
9/20
Documentation
9/20
Setup experience
10/20
Why each score
Reliability 9/20
README and manifest claim MCP API, REST API and sandbox tooling, but no executable static evidence exists: no visible CI workflows, no committed tests, no tool-list-to-implementation comparison. Quick-start examples cover the happy path (install, upload, query), but error handling, edge cases and dependency-failure diagnostics are absent. Per static calibration, capped at 12; awarded 9 for a plausible happy path with zero verification evidence.
Security and permissions 8/20
No red-line findings: no malware indicators, no real tokens, no covert exfiltration; the API key is correctly flagged as optional secret in the manifest. But significant deductions: auth is optional, so the local service is reachable by any client by default; guardrails/security are marketing claims without visible least-privilege, dangerous-operation confirmation, or data-boundary implementation evidence; a SQL sandbox with broken guardrails could reach the underlying filesystem. Scored 8 — not a red line, but security boundaries lack evidence.
Maintenance 9/20
MIT license, clear repository ownership, non-archived, versioned at v1.3.0 with OCI and Homebrew distribution, indicating basic release discipline. However, low stars (21), 11 open issues, and no static evidence of issue responsiveness, dependency update cadence, or a security-response channel. Awarded 9.
Documentation 9/20
README gives clear examples for install, startup, upload, query, Claude Code integration and the Explorer UI, and links to full docs (DOCS.md). But only the README itself is reviewable: API key usage, tool parameters, limits, cost and troubleshooting are not demonstrated here, and the external doc content is unverifiable. Awarded 9.
Setup experience 10/20
Install path is clear: one Homebrew command, one binary start, curl verification, and a concrete claude mcp add client config with npx mcp-remote. Deductions: the manifest declares stdio transport while the README uses an HTTP stream endpoint (mcp-remote http://localhost:8081/stream) — an inconsistency users must resolve themselves; and without CI+test evidence, static calibration caps setup at 15. Awarded 10.

Static review · not runListed 2026-09-08

Read the FMRS scoring method →

Fit and risk

What it can accessUses the networkConnects to a database

Best for

  • AI developers needing a secure, isolated data environment
  • Teams wanting controlled data exploration and analysis
  • Users wanting to quickly try DuckDB capabilities

Not for

  • Production environments requiring long-term persistent storage
  • Scenarios needing high concurrency or complex data pipelines
  • Users looking for a full-fledged database management tool

Required permissions

  • Can read user-provided CSV files
  • Can execute SQL queries (only within sandbox)
  • Can access data via API or MCP

Risks and side effects

  • Data may be ephemeral; save promptly
  • Sandbox may not directly access production databases
  • Relies on external components like Docker

Setup

Before you start

Runtime：Prebuilt binary · Node.js

  1. Install and start via Homebrew: brew tap aliengiraffe/spaceship && brew install spotdb, then run spotdb to start the server.
  2. Add MCP via Claude Code: claude mcp add spotdb -s user -- npx -y mcp-remote http://localhost:8081/stream.
  3. Upload a CSV and query data as per documentation.

Check that it works

Run `spotdb` to start the server, then add it in Claude Code with `claude mcp add spotdb -s user -- npx -y mcp-remote http://localhost:8081/stream`; after restarting the client, confirm SpotDB's MCP tools appear in the tool list, or open http://localhost:8080/explorer to verify the Explorer UI loads.

Troubleshooting

  1. Ensure SpotDB server is started: `spotdb`
  2. Check ports 8080 and 8081 are open
  3. Verify CSV upload format includes table name and file path

Things to try

Once connected, you can ask your AI assistant things like:

  • Upload data.csv to SpotDB as mytable and show the first 10 rows
  • Run SELECT * FROM mytable LIMIT 10 against the uploaded CSV
  • Take a snapshot of the current sandbox data
  • Use SpotDB to count the rows in mytable with SQL

Use cases

Let AI agents analyze data in a safe sandbox without touching production databases
Quickly upload CSVs and run SQL queries for data exploration
Integrate an ephemeral data environment into MCP or REST workflows

Overview

SpotDB is a lightweight, ephemeral data sandbox designed for large language models (LLMs) and agentic workflows. It provides a secure, isolated environment that allows AI agents and scripts to analyze data without direct access to production databases. This prevents accidental data modification, ensures data privacy, and enforces guardrails for safe data exploration. It supports MCP API for seamless AI integration and a REST API for traditional systems.

Similar servers

Compare with other MCP servers

Source revision cfbbef27f89d Data synced 2026-10-19 Read the FMRS scoring method