Sqbyl MCP Server

The full sqbyl dev toolkit: introspect, profile, annotate, synth, eval, Coach, judges, console, optimizer, release builder.

Local serverstdioPython

What is the Sqbyl MCP Server?

The full sqbyl dev toolkit: introspect, profile, annotate, synth, eval, Coach, judges, console, optimizer, release builder.

What you get

Bring your own database and one LLM provider key (Anthropic or OpenAI). sqbyl uses your chosen model to both answer natural-language questions against your data and coach you on how to make the agent answer them better — then ships the result as a single portable file you can drop into production.

  • No black box. — Every prompt, judge, and improvement proposal is readable, editable plain text/JSON
  • No second vendor. — A single provider key (Anthropic or OpenAI) powers the agent, the judges, and the Coach. Context selection is LLM/lexical, so there's no embeddings provider or vector store to run
  • No surprise bill. — The free, deterministic work (connect, profile, infer joins) runs first at $0. Paid work is estimated up front, metered live, and capped by --budget
  • Versioned like code. — Your whole "agent" is a directory of YAML you diff, review, and git revert
  • Defensible by design. — The headline accuracy is deterministic and measured on a held-out set the improvement loop can never touch — so "we hit 94%" is a claim that survives scrutiny, not a benchmark you overfit. (more below)

Configuration and credentials

  • Python (managed via uv) - An LLM provider API key — Anthropic (ANTHROPIC_API_KEY) or OpenAI (OPENAI_API_KEY) - A SQL database, reachable read-only (DATABASE_URL). DuckDB and Postgres are the first-class dialects; SQLite, MySQL, Snowflake, and BigQuery are supported behind a dialect seam. ---

More MCP options

Compare MCP servers by category, or browse hosted MCP servers if you prefer an endpoint over a local process.

Maintainer setup documentation.

How to install the Sqbyl MCP Server

{
  "mcpServers": {
    "sqbyl": {
      "command": "uvx",
      "args": ["sqbyl"],
      "env": {
        "ANTHROPIC_API_KEY": "your-value",
        "DATABASE_URL": "your-value",
        "OPENAI_API_KEY": "your-value"
      }
    }
  }
}

Add to claude_desktop_config.json, then restart Claude Desktop.

Configuration

  • Python (managed via uv) - An LLM provider API key — Anthropic (ANTHROPIC_API_KEY) or OpenAI (OPENAI_API_KEY) - A SQL database, reachable read-only (DATABASE_URL). DuckDB and Postgres are the first-class dialects; SQLite, MySQL, Snowflake, and BigQuery are supported behind a dialect seam. ---
VariableDescriptionRequired
ANTHROPIC_API_KEYCredential the server authenticates with.Yes
DATABASE_URLEndpoint or connection string the server talks to.Yes
OPENAI_API_KEYCredential the server authenticates with.Yes

Frequently asked questions

The full sqbyl dev toolkit: introspect, profile, annotate, synth, eval, Coach, judges, console, optimizer, release builder. Confirm the tools offered by your installed version in the connected client. Tool availability and permissions depend on the server and account configuration.