dolt

Overview

The dolt-mcp-server is a Model Context Protocol (MCP) server that allows AI assistants and agents to interact directly with Dolt, a SQL database with built-in Git-style version control. It enables AI-driven workflows to query data, inspect schema changes, create and manage branches, review diffs, and reason about historical data — all without switching tools or manually using Dolt’s CLI.

This server is especially useful for data analysis, auditing, experimentation, and collaborative workflows where tracking and understanding data changes over time is critical.

Transport

streamable-http

Tools

• add_dolt_remote
• add_dolt_test
• alter_table
• clone_database
• create_database
• create_dolt_branch
• create_dolt_commit
• create_table
• delete_dolt_branch
• describe_table
• dolt_fetch_all_branches
• dolt_fetch_branch
• dolt_pull_branch
• dolt_push_branch
• dolt_reset_hard
• dolt_reset_soft
• drop_database
• drop_table
• exec
• list_databases
• list_dolt_branches
• list_dolt_commits
• list_dolt_remotes
• merge_dolt_branch
• move_dolt_branch
• query
• remove_dolt_remote
• remove_dolt_test
• run_dolt_tests
• select_active_branch
• select-version
• show_create_table
• show_tables
• unstage_all_tables
• unstage_table

Key Capabilities

• Versioned data access — Query data at specific commits, branches, or points in time.
• Git-style workflows for data — Create branches, commit changes, and inspect diffs programmatically.
• Schema and data auditing — Understand how tables and records have evolved over time.
• Safe experimentation — Explore or modify data in isolated branches without affecting production state.
• SQL-native interaction — Use familiar SQL queries while benefiting from built-in version control.

How It Works

The dolt-mcp-server runs as a local or containerized MCP service connected to one or more Dolt repositories. Once configured, it exposes Dolt’s SQL engine and version-control operations as MCP tools that AI clients can invoke.

When an AI assistant calls a tool, the server runs the SQL query against the current branch or a specified revision. Versioning tools such as diff, checkout_branch, and commit allow agents to reason about changes, compare data across branches, and manage evolution over time. All results are returned in structured form over the MCP protocol.

By abstracting Dolt’s CLI and internal mechanics, the server allows AI assistants to treat versioned data as a first-class concept. This enables workflows like “compare last week’s data to today,” “explain what changed in this table,” or “create a branch to test a new dataset” entirely through conversational and automated AI interactions.