dbt Cloud Catalog
Great if you run dbt Cloud. Not available if you run dbt Core on premises. Taavik fills that gap with a tool that observes the actual database, not the dbt model.
Database documentation tool
A database documentation tool that does not go stale.
Most database wikis are wrong within a week of being written, and everyone on the team quietly knows it. Taavik makes the documentation a function of the schema instead of a chore somebody owes: the agent scans, the diff rewrites the page, and the change history records when it happened. Whatever you wrote by hand stays exactly where you put it.
What teams need from a database documentation tool
Three jobs the wiki cannot do anymore.
Confluence, Notion, dbDocs, an ER diagram in Lucidchart. Different tools, same weak point: every one of them waits for a human to remember. After the third missed migration people stop trusting what the page says, and once trust is gone the page is dead whether or not anyone deletes it.
Auto-update from the schema
A scan picks up a new column and the page shows it. A column disappears and the page drops it, with the change history keeping the date it happened. Nobody synchronises anything.
Preserve manual notes
The technical metadata is generated. The prose paragraphs are yours. Auto-updates touch the metadata only and never overwrite a description.
Built-in change history
Every scan produces a diff. The doc shows the most recent changes per object, so a new hire reads the present and the recent past in one place.
Works on private databases
The agent runs on a host inside your network and the cloud receives the catalog, which is usually enough for compliance to stop asking difficult questions.
Where Taavik fits
If you tried these and they did not work.
DataHub or OpenMetadata self-hosted
Powerful, but require half a person to operate. Taavik is a single agent and a hosted control plane that keeps documentation alive without dedicated operators.
dbDocs and ER diagram exports
Static snapshots. Useful for a one-shot deliverable, painful as living documentation. Taavik refreshes on every scan and never asks you to re-run the export.
Confluence and Notion pages
Easy to start, hard to keep current. Taavik moves the technical metadata to a tool that updates itself. Your wiki keeps the high-level prose; Taavik keeps the schema accurate.
What is supported
PostgreSQL, SQL Server, MySQL, Amazon Redshift.
One workspace, many connections, many engines. Provider quirks normalized at scan time, the documentation reads consistently across databases.
Database engines
Documentation pages for PostgreSQL, SQL Server, and MySQL.
PostgreSQL documentation tool
Generate PostgreSQL documentation from catalogs and keep schema changes visible.
Read provider pageSQL Server documentation tool
Document SQL Server metadata, notes, and change history from private scans.
Read provider pageMySQL documentation tool
Keep MySQL documentation current with metadata scans and preserved manual notes.
Read provider pageGuides
Build a documentation workflow that stays current.
Living database documentation
How generated metadata, manual notes, and schema changes become a documentation layer the team can trust.
Read the guidePrivate database documentation
Why private databases need an agent-based scan instead of direct cloud database access.
Read the guideKeep database docs up to date
A practical operating model for scans, change review, generated facts, and preserved manual notes.
Read the guideThe engine behind the docs
Documentation that refreshes itself is a change-detection problem first.
The documentation stays useful only because the diff engine underneath knows when something moved, when something appeared, and when a column that looked new is actually the old one under a different name. Two pieces on the engine describe how that layer works, and why the team can trust the rendered page after the twelfth scan.
Schema drift, and why the wiki loses first
The gap between the schema the migration folder describes and the schema the running database actually has. Where it opens, why "we have migrations" does not close it, and what a diff engine needs to make that gap visible before the next bug.
Read the pieceDatabase observability, but for structure
Query performance monitoring watches how the database behaves at runtime. That is not the same signal as watching how the schema itself changes over time. Why structural change deserves to be a first-class observability layer alongside the runtime one, and what an implementation of it looks like.
Read the pieceCompare
Compare focused documentation workflows.
Dataedo alternative
Compare broad data catalog workflows with focused private database documentation and operational SQL.
Read comparisonDbSchema alternative
Compare visual design workflows with living documentation refreshed from database scans.
Read comparisondbdocs alternative
Compare DBML docs-as-code with scan-based documentation for private databases.
Read comparisonFAQ
Database documentation tool for PostgreSQL, MySQL and SQL Server
Which databases are supported? +
PostgreSQL, MySQL, SQL Server, and Redshift. The on-premise agent connects to any of them and reads schema metadata through the standard information schema.
Is this a database documentation generator or a database documentation tool? +
Both, and the difference matters less than the vendor page usually suggests. A generator runs once and produces a static artifact; a tool refreshes on a schedule and lets the team edit around what it produces. Taavik works as a tool: every scan is a fresh generation, and the manual notes ride along across generations without being overwritten.
Is there a free database documentation software plan? +
Yes. One database connection, one workspace member, unlimited scans, thirty days of schema change history. Enough for a small team to keep a single production database documented end to end without a card on file.
How does the tool keep MySQL, PostgreSQL, and SQL Server documentation consistent across engines? +
Each provider exposes metadata through its own catalogs (information_schema on MySQL, pg_catalog on PostgreSQL, sys.tables and friends on SQL Server). The agent normalizes those into a shared shape at scan time, so the rendered documentation reads the same regardless of the engine. Provider-specific fields that have no cross-engine equivalent are labeled explicitly.
Does the database schema documentation tool detect renames or does it show them as drop plus add? +
It detects renames. A naive diff would surface a rename as a drop plus an add and double the noise. The scan compares shape (data type, nullability, ordinal, name similarity) between the missing object and the new one, and when the match is strong enough it emits a single rename event with a score. The piece on schema drift detection goes into the mechanism in detail.
Where is my data stored? +
The control plane runs on dedicated virtual machines inside the European Union. What it holds is the catalog, your manual notes and the change history. The rows in your tables stay in your network.
Is the documentation hosted on Taavik or self-hosted? +
The documentation is rendered in the workspace, served by the cloud control plane. The data behind it is catalog metadata, manual notes, and change history. Credentials stay local.
Can I edit the documentation manually? +
Yes. Add descriptions to tables, paragraphs to columns, known-issue blocks to objects. Auto-updates touch only the technical metadata; your text stays untouched.
Can an AI assistant read the documentation? +
Yes. The workspace exposes an MCP connector at https://api.taavik.com/mcp. Any assistant that speaks MCP, Claude and Claude Code among them, can read the catalog through it. The connector is read-only and never receives the database credentials. The MCP hub page covers the boundary in detail.
Does Taavik replace dbt docs? +
Taavik documents the database as it is. dbt docs documents the dbt models you wrote. They serve different audiences. Many teams use both.
Try it
Generate the first documentation in under ten minutes.
Free for one connection, forever. The agent is small enough that the first scan is usually done before you have finished reading the setup page.