> For the complete documentation index, see [llms.txt](https://docs.keeper.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.keeper.io/release-notes/desktop/keeperdb/keeperdb-3.0.0.md).

# KeeperDB 3.0.1

Released on Nov 5, 2026

### What is KeeperDB

**KeeperDB** is a fast, secure, cross-platform database management tool. Use it inside KeeperPAM connections or as a standalone desktop app on Windows, macOS, and Linux.

{% embed url="<https://vimeo.com/1220632470?share=copy&fl=sv&fe=ci>" %}

Query, explore, and operate PostgreSQL, MySQL, SQLite, Microsoft SQL Server, Oracle, Amazon Redshift, MongoDB, and Amazon DynamoDB from one interface.

KeeperDB is built for engineers and data scientists. It replaces legacy tools like DBeaver, MySQL Workbench, and pgAdmin. In KeeperPAM, it brings core database workflows into a fully managed passwordless experience.

Quick Links: [Product Documentation](/keeperpam/privileged-access-manager/keeperdb.md) | [Download Now](https://www.keepersecurity.com/download.html?t=db)

### Version 3.0.1 Summary

KeeperDB 3.0.1 brings the new KeeperDB MCP server. AI agents like Claude Desktop and Claude Code can now explore your databases (read-only), and KeeperDB controls exactly which connections and tools they can use. The desktop app now allows you to sign in with your Keeper account, and KeeperPAM subscribers unlock premium features, including Broadcast Mode and the MCP server.

This release also lets KeeperAI work with the AI provider of your choice and ask before it runs any statement that changes data, adds mongosh syntax for MongoDB, makes signing in on a new device smoother, and includes important security fixes for the desktop app and the Docker image.

{% hint style="info" %}
KeeperDB is available standalone, and built into the upcoming Keeper Gateway 1.9 release.
{% endhint %}

### KeeperDB MCP Server

KeeperDB 3.0 includes an MCP server, so AI agents like Claude Desktop and Claude Code can explore your databases through the Model Context Protocol. The agent talks to your running KeeperDB desktop app and never sees a database password: KeeperDB enforces read-only access, the connection allowlist, and row limits itself on every request.

Ask an AI agent about your schema and data in plain language

<figure><img src="https://1549319098-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LSGErxSfzX6xZALob42%2Fuploads%2F3xSQ1EPALv0EdvtRJnSz%2F01-mcp-claude.png?alt=media&amp;token=3a971558-7a83-4632-b3ec-cf93b12e8917" alt=""><figcaption></figcaption></figure>

* One-click install for Claude Desktop, plus copy-paste setup instructions for Claude Code and any other MCP client
* Read-only by design: on PostgreSQL, MySQL, and SQLite, MCP sessions are also read-only at the database itself
* Agents only see the saved connections you check in the connection allowlist
* Large results are paged and size-limited, so a single query can't flood the agent
* The MCP Activity log records every request, including denied ones, and keeps that history across restarts

Set up an MCP client from Settings → MCP, and choose which connections and tools agents can use

<figure><img src="https://1549319098-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LSGErxSfzX6xZALob42%2Fuploads%2Fxk8o2Y1U72TNFeHnA68w%2F02-mcp-allowlist.png?alt=media&amp;token=86c4da4c-347b-4524-850f-49534ca18457" alt=""><figcaption></figcaption></figure>

Installing the Claude Desktop extension grants no access on its own

<figure><img src="https://1549319098-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LSGErxSfzX6xZALob42%2Fuploads%2FIjQ7AaZMCWe8zQcI6v7G%2F02a-mcp-install.png?alt=media&amp;token=f444559f-020d-41ba-b6a1-6a70a4773454" alt=""><figcaption></figcaption></figure>

Each new client must be approved in KeeperDB before it can connect. The approval shows exactly what the client will be able to reach, and you can revoke it at any time.

<figure><img src="https://1549319098-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LSGErxSfzX6xZALob42%2Fuploads%2FUSByK3bFls9MtmQHHbVO%2F02b-mcp-pairing-approval.png?alt=media&amp;token=c63948d8-84ee-4140-b686-5ad1e4b37ad2" alt=""><figcaption></figcaption></figure>

{% hint style="info" %}
The MCP server is a premium feature of the KeeperDB desktop app. It isn't available in KeeperPAM connections launched through Keeper Gateway.
{% endhint %}

### KeeperAI

KeeperAI now works with the AI provider of your choice, and it asks for your approval before it runs anything that changes your data.

Choose from OpenAI, Anthropic, Google Gemini, Azure OpenAI, AWS Bedrock, Google Vertex AI, or any OpenAI-compatible endpoint, such as Ollama or vLLM. Keeper AI Gateway is coming soon.

<figure><img src="https://1549319098-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LSGErxSfzX6xZALob42%2Fuploads%2FgSBLeUtlrw2Ilmu0gxeV%2F06-ai-providers.png?alt=media&amp;token=1062f1f7-163e-4611-90b1-d568588b60d9" alt=""><figcaption></figcaption></figure>

* API keys are stored in your operating system's keychain, never on disk in plaintext
* A saved API key is tied to its provider endpoint, so it is never sent to a different host
* KeeperAI asks for your approval before it runs a statement that changes data or the schema

KeeperAI shows the exact statement and flags risky ones, such as an UPDATE with no WHERE clause, before anything runs.

<figure><img src="https://1549319098-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LSGErxSfzX6xZALob42%2Fuploads%2FMrWrVeFU4yJJo9j1yI7r%2F07-ai-approval.png?alt=media&amp;token=06cb3e77-b528-476f-ae62-d730ba0d7021" alt=""><figcaption></figcaption></figure>

### Keeper Account Sign-In and Premium Features

The KeeperDB desktop app now signs in with your Keeper account. KeeperPAM subscribers unlock premium features, starting with Broadcast Mode and the MCP server.

* Sign in with your master password or SSO
* Two-factor authentication supports a choice of method, including security keys such as YubiKey
* New-device approval explains each option clearly: approve from Keeper Vault or the mobile app, use the email link, or enter an approval code
* Optional "Stay signed in" keeps you signed in across restarts and upgrades
* Free users can see what's included from the Premium Features panel

### MongoDB

MongoDB queries can now be written in mongosh syntax, the same shell syntax used in MongoDB's own tools.

Run mongosh-style queries in the Query tab and browse the results as documents

<figure><img src="https://1549319098-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LSGErxSfzX6xZALob42%2Fuploads%2F6jAxCcogUVsfPr6mb1Vi%2F08-mongosh.png?alt=media&amp;token=be83377e-12c0-4930-bd9d-80a7c515ca5c" alt=""><figcaption></figcaption></figure>

* "Use TLS" is checked by default for new MongoDB connections
* The database selected in the sidebar now applies to the Query tab

### Connections and Usability

Save a connection directly from the New Tab connection form, without going through Settings

<figure><img src="https://1549319098-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-LSGErxSfzX6xZALob42%2Fuploads%2F9y4YPeVFxzkX1unxnAQr%2F09-save-connection.png?alt=media&amp;token=69f2205a-cac6-487c-8155-2d6310ca7323" alt=""><figcaption></figcaption></figure>

* SQL Server: choose the authentication mode when opening an additional connection, and cancel an Entra ID sign-in that is in progress
* Azure SQL databases that are paused now resume and connect, instead of failing on the first attempt
* Query results cut off at the row limit are now flagged as truncated, so you can page past them
* Keyboard shortcuts can now use the plus (+) key
* PostgreSQL and Amazon Redshift connections with no database set, such as a KeeperPAM record without a default database, now connect to the default `postgres` or `dev` database
* Oracle: the Native driver no longer writes diagnostic (ADR) files, and the connection form recommends the Thin driver for KeeperPAM Gateway connections
* The theme and color preset are kept after you quit the macOS app
* Numeric settings, such as Connect Timeout, save when you press Enter or Tab or leave the field

### Security and Stability

* The desktop app's local API now requires the app's own access token on every request, so other programs on your computer can't use it to open saved connections or change the app
* DNS-rebinding protection: the KeeperDB server only accepts requests addressed to localhost, an IP address, or a host name listed in `KEEPERDB_ALLOWED_HOSTS`, and the desktop app accepts local requests only
* The KeeperDB Docker image no longer includes two distrusted root certificates (AAA Certificate Services and emSign ECC Root CA - G3)
* Credentials are kept out of the audit log and application logs, including Oracle `user/password@` connect strings
* Only one copy of the desktop app can run at a time
* Windows: KeeperDB now unlocks correctly when launched elevated through Keeper Endpoint Privilege Manager (KEPM)

### Known Issues

* If a new device is approved more than about 55 seconds after the approval prompt appears, sign-in can end with "Too many login attempts". Click **Sign in again** and approve within about 55 seconds. A fix is planned for the next release.
* Oracle Thin connections show at most the first 100 rows of a query result, without marking the result as cut off. Use the Native driver when you need more rows.
* After KeeperDB restarts, an MCP client's first request may report that KeeperDB isn't running. Ask again, and if it keeps failing, restart the MCP server in your client.

### Resources

* [KeeperDB Documentation](/keeperpam/privileged-access-manager/keeperdb.md)
* [KeeperDB Proxy Documentation](/keeperpam/privileged-access-manager/keeperdb-proxy.md)
* [KeeperDB Feature Page](https://www.keepersecurity.com/features/keeper-db/)
* [KeeperAI Documentation](/keeperpam/privileged-access-manager/keeperai.md)
* [KeeperPAM](https://www.keepersecurity.com/privileged-access-management/)

### Roadmap

We publish monthly updates based on customer feedback. Send feature requests and bug reports to <pam@keepersecurity.com>, or post on our [Reddit community page](https://www.reddit.com/r/KeeperSecurity/).


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.keeper.io/release-notes/desktop/keeperdb/keeperdb-3.0.0.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.