terrabase
LLM-assisted, safety-gated Postgres migrations exposed as an MCP server, using a deterministic rule engine over Postgres's own parser AST for safety enforcement, with two-phase approval and append-only audit ledger.
README
terrabase-v2
LLM-assisted, safety-gated Postgres migrations, exposed as an MCP server.
A local LLM may draft migration SQL, but a deterministic rule engine over Postgres's own parser AST is the sole safety authority. Every write goes through a two-phase, approval-gated flow and lands in an append-only audit ledger.
The LLM proposes, the rules decide.
Layout (pnpm monorepo)
| Package | Purpose |
|---|---|
packages/core |
shared types, zod schemas, errors, config, DSN redaction |
packages/pg-ast |
libpg_query (WASM) parse + traversal + typed node guards |
packages/pg-safety |
the 30-rule deterministic safety analyzer (the heart) |
packages/pg-introspect |
live schema → compact JSON model |
packages/pg-advisor |
EXPLAIN (FORMAT JSON) parsing + HypoPG index advice |
packages/pg-migrate |
append-only audit ledger + transactional, approval-gated apply |
packages/llm |
Ollama structured output + retry-on-zod-error |
apps/mcp-server |
the 5 MCP tools + two-phase orchestration |
Full design: docs/specs/2026-07-07-terrabase-v2-design.md.
Quick start
pnpm install
pnpm build # tsup builds every package
pnpm typecheck # tsc --noEmit, strict
pnpm test # vitest unit tests (run everywhere)
Configure via env (see .env.example): TERRABASE_DATABASE_URL, OLLAMA_HOST,
OLLAMA_MODEL, TERRABASE_PG_VERSION.
Run the MCP server
pnpm mcp # speaks MCP over stdio
Register it with any MCP client (command: node apps/mcp-server/dist/index.js).
Graceful degradation
The parser + rule engine run with no external services. If Ollama is down, pass sql
directly to propose_migration. If there's no database, SQL is still parsed and analyzed
offline against TERRABASE_PG_VERSION. Docker-backed integration tests skip automatically when
Docker is unavailable.
MCP tools
inspect_schema— compact schema model (read-only)propose_migration—intent→LLM-drafted SQL or rawsql→ analyze → returns aSafetyReport+proposalId(applies nothing)apply_migration—proposalId+approvalToken→ gated, transactional apply → ledgerexplain_query— parsed plan tree + flagged problemsadvise_index— HypoPG-simulated index recommendations (Tier-1 catalog fallback)
Desktop app (Strata)
An Electron GUI over the same engine lives in apps/desktop. It runs the
parser + rule engine locally, auto-discovers project databases, and renders the
SafetyReport visually.
pnpm --filter @terrabase-v2/desktop dev
Contributing
Contributions are welcome! Please read CONTRIBUTING.md for setup, the PR checklist, and the one non-negotiable rule (the LLM is advisory only). All participants are expected to follow our Code of Conduct.
Security
Found a vulnerability? Please report it privately — see SECURITY.md. Do not open a public issue for security reports.
License
Licensed under the Apache License 2.0. © 2026 Terrabase.in.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.