v0.8.0 Released - AI Linter & MCP Exports

AI Database Schema Generator & Context Engine.

Generate AI-optimized database documentation, LangChain tools, MCP context, schema diffs, and LLM-ready exports.

Git for AI Database Context.

Get License Key Install Now

bash

$ uv run schemap generate --format=ai

NATIVE EXPORTS FOR

from langchain.tools import BaseTool
from pydantic import BaseModel

class DatabaseSchemaTool(BaseTool):
    name = "database_schema_context"
    description = """
    Table: users
    Columns: id, email, created_at
    Relationships: users -> orders (1:N)
    """
    
    def _run(self, query: str) -> str:
        # Injects pristine context into agent
        return self.description

{
  "name": "schemap_server",
  "version": "1.0.0",
  "tools": [
    {
      "name": "get_database_schema",
      "description": "Retrieves highly compressed AI context.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "query": { "type": "string" }
        }
      }
    }
  ]
}

{
  "tables": [
    {
      "name": "users",
      "columns": [
        { "name": "id", "type": "UUID", "pk": true },
        { "name": "email", "type": "VARCHAR" }
      ]
    }
  ],
  "relationships": [
    { "source": "users", "target": "orders" }
  ]
}

Don't Waste 50,000 Tokens on Raw SQL.

Frontier models reward dense structure, not conversational noise. Schemap translates massive SQL dumps into perfect AI context.

Before (Raw SQL Dump)

CREATE TABLE users (
  id UUID PRIMARY KEY,
  email VARCHAR(255) NOT NULL,
  created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
  password_hash VARCHAR(255),
  stripe_cust_id VARCHAR(50)
);

CREATE TABLE orders (
  id UUID PRIMARY KEY,
  user_id UUID,
  amount DECIMAL(10,2),
  status VARCHAR(20),
  FOREIGN KEY (user_id) REFERENCES users(id)
);
-- ... 50,000 tokens later

After (Schemap AI Output)

TABLE users [Customer]
PK id [Identifier]
COL email string [Email]

TABLE orders [Invoice]
PK id [Identifier]
COL amount float [Amount]

REL users 1:N orders

JOURNEYS
PATH users -> orders

Core Infrastructure for AI Agents

Everything you need to inject live database schema context into your agentic pipelines.

AI Context Compression

52k → 3k tokens

We strip out conversational noise and inject pure, heavily-minified structural DSL that keeps LLM attention heads focused.

Native Agent Exports

LangChain • LlamaIndex • MCP

Generate highly contextual, directly-executable Python tools pre-loaded with your database schema for immediate use.

The AI Schema Linter

Score: 84/100

Is your schema hurting LLM accuracy? Run `schemap score` to catch missing Primary Keys, detached tables, and ambiguous column names.

Schema Diff Intelligence

+ invoices, - users.age

A true version-control system. Catch silent schema drifts in CI/CD pipelines before they break your downstream text-to-SQL agents.

Built for the Modern Data Stack

For AI Engineers

RAG Data Systems
Text-to-SQL Pipelines
Autonomous Agents

For Data Teams

Living Documentation
Schema Governance
Database Onboarding

For Startups

Slash Token Costs
Improve AI Accuracy
Ship Faster

Install in 2 Seconds

Schemap leverages `uv` to maintain a zero-friction footprint.

pip install schemap-tool

uv run schemap init

This generates your schemap.yaml. Update it with your database connection details:

                        database:
  connection_url: "postgresql://user:password@localhost:5432/my_db"
  exclude_tables:
    - "spatial_ref_sys"

domain:
  name: "ecommerce"

license_key: "your-license-key"
                    

Frictionless Licensing

Simple, lifetime value. No heavy web logins required.

Free

For developers building local side projects.

Maximum 50 database tables
Local execution only
Schema linting & basic exports
Token footprint metrics