Dynoxide

Name: Dynoxide
Author: nubo-db

Local DynamoDB emulator on SQLite whose MCP mode gives coding agents 35 table and item tools

By nubo-db Rust Local · stdio Databases Developer tools Testing

GitHub npm Website

62 popularity score ★ 105 ↓ 16k/wk

Connect your client

Local (npm)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server.

Or add it by command or config

Claude CodeTerminal

claude mcp add dynoxide -- npx -y dynoxide mcp

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "dynoxide": {
      "command": "npx",
      "args": [
        "-y",
        "dynoxide",
        "mcp"
      ]
    }
  }
}

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "dynoxide": {
      "command": "npx",
      "args": [
        "-y",
        "dynoxide",
        "mcp"
      ]
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "dynoxide": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "dynoxide",
        "mcp"
      ]
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.dynoxide]
command = "npx"
args = ["-y", "dynoxide", "mcp"]

Gemini CLITerminal

gemini mcp add --scope user dynoxide npx -- -y dynoxide mcp

Local (container)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server.

Or add it by command or config

Claude CodeTerminal

claude mcp add dynoxide -- docker run -i --rm ghcr.io/nubo-db/dynoxide mcp

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "dynoxide": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/nubo-db/dynoxide",
        "mcp"
      ]
    }
  }
}

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "dynoxide": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/nubo-db/dynoxide",
        "mcp"
      ]
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "dynoxide": {
      "type": "stdio",
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/nubo-db/dynoxide",
        "mcp"
      ]
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.dynoxide]
command = "docker"
args = ["run", "-i", "--rm", "ghcr.io/nubo-db/dynoxide", "mcp"]

Gemini CLITerminal

gemini mcp add --scope user dynoxide docker -- run -i --rm ghcr.io/nubo-db/dynoxide mcp

Settings 1

`DYNOXIDE_MCP_AUTH_TOKEN`	Optional: Bearer token for the --http transport, used instead of the generated, persisted one; needed for a non-loopback bind. Not used over stdio.

In a container, a setting goes in with -e NAME=value after docker run.

What it does

Dynoxide is a stand-in for Amazon DynamoDB that runs on your own machine. It keeps data in SQLite and ships as one native binary, so it needs no JVM or Docker. The same binary can serve the DynamoDB HTTP API for SDKs and tests, or, with the `mcp` subcommand, run as an MCP server so a coding agent can create tables, write and read items, run queries, scans, PartiQL and transactions, and read stream records against a local copy instead of a real AWS account.

Without `--db-path` the database lives in memory and is gone when the process stops; with it, the data is kept in a SQLite file. Agent-oriented extras include named snapshots you can roll back to, an automatic snapshot before every `delete_table` (the ten most recent are kept), a `--read-only` switch that refuses every write tool, caps on how much a query or scan returns, and a `--data-model` flag that loads a OneTable schema so the agent sees entity names and key templates. `dynoxide import --mcp` can also load a DynamoDB table export into memory, optionally anonymised, and serve that copy.

The v1.3.0 source registers exactly 35 tools (src/mcp/server.rs), matching the docs' count. In the code, `--max-items` and `--max-size-bytes` have no default, so query and scan results are not capped unless you set them. The server talks over stdio by default; `--http` switches to Streamable HTTP on port 19280, bound to 127.0.0.1 unless `--host` is given.

What it needs

Runtime	Node.jsorDocker

The prebuilt binary for macOS (x64, arm64), Linux (x64, arm64) or Windows (x64). The npm way needs Node.js with npx; the package pulls the right binary through optionalDependencies. The container way needs Docker. No AWS account or credentials are needed.

Limits

Built for local development, testing and CI, not as a production DynamoDB. Cloud-only operations (backups and point-in-time restore, global tables, Kinesis streaming, resource policies, capacity management) return UnknownOperationException. ConsistentRead is accepted but does nothing, streams expose a single shard, and transaction-contention errors are not emulated. Vector search is exact brute-force KNN and differs from AWS in ways that can make a test pass locally and fail on AWS; an index added via `update_table` is not searchable at once. The `--data-model` schema is context only: writes are not checked against it. The WebAssembly build is still missing several operations.

Who it suits

Developers building on DynamoDB who want an agent to design tables, seed data and try queries without touching a real AWS account or paying for it. It runs entirely on your machine with no sign-in. It does not reach AWS at all, so it is the wrong pick for inspecting or changing live DynamoDB tables; use a server that works with your AWS account for that. For working on a copy of real data, pair `dynoxide import --mcp` with `--mcp-read-only`.

Tools 35

list_tables: List the tables in the local database
describe_table: Show a table's keys, indexes, vector indexes and status
create_table: Create a table, optionally with GSIs, LSIs, streams or vector indexes
delete_table: Drop a table, taking an automatic snapshot first
update_table: Change a table's settings, indexes or vector indexes
get_item: Fetch one item by its key
put_item: Write or replace one item, with an optional condition
update_item: Change attributes of one item with an update expression
delete_item: Remove one item by its key
batch_get_item: Fetch many items across tables in one call
batch_write_item: Put or delete many items in one call
bulk_put_items: Load a large set of items into a table
query: Read items by partition key, on a table or an index
scan: Read every item in a table or index, with filters
search_vectors: Run a nearest-neighbour search on a vector index
transact_get_items: Read several items as one transaction
transact_write_items: Write several items all-or-nothing
execute_partiql: Run one PartiQL SELECT, INSERT, UPDATE or DELETE statement
batch_execute_partiql: Run a batch of PartiQL statements
execute_transaction_partiql: Run PartiQL statements as one transaction
update_time_to_live: Turn TTL on or off for a table
describe_time_to_live: Show a table's TTL setting
sweep_ttl: Delete items whose TTL has passed, on demand
tag_resource: Add tags to a table
untag_resource: Remove tags from a table
list_tags_of_resource: List a table's tags
list_streams: List the tables' change streams
describe_stream: Show a stream and its single shard
get_shard_iterator: Get a position for reading a stream shard
get_records: Read change records from a stream
create_snapshot: Save a point-in-time copy of the database
restore_snapshot: Roll the database back to a saved snapshot
list_snapshots: List the saved snapshots
delete_snapshot: Remove a saved snapshot
get_database_info: Report database details, including any import notices

Popularity

62 popularity score, on our list (the bar is 36)

★ 105 ↓ 16k/wk

GitHub stars	105	24.3 points
Stars in 30 days	+7	4.5 points
npm downloads, last week	15,696	33.6 points

Measured 2026-11-07. How the score works.

Where it is published

Code	nubo-db/dynoxide ★ 105
npm	dynoxide 1.5.0, 2026-10-23, the one the command runs
Container	ghcr.io/nubo-db/dynoxide

What it can reach

Runs	On your machine
Reaches	A database, Files on your machine
Changes things	Yes: it can write, send, create or delete
Sign-in	None
Licence	(MIT OR Apache-2.0) (npm)
Last code change	2026-11-05

Runs entirely on your machine; it makes no calls to AWS and needs no AWS credentials. Its tools read and change only Dynoxide's own database: in memory by default, or the SQLite file given with `--db-path`, plus snapshot copies of it. Most tools can create, change or delete tables and items; `--read-only` makes every write tool refuse and limits PartiQL to single SELECT statements. Over stdio there is no sign-in. The `--http` transport needs a bearer token on every request: on a loopback bind one is generated and saved in a per-user config file; a non-loopback bind will not start without an explicit token, and `--no-auth` works on loopback only. An anonymised import may still hold original values if exposures were accepted; `get_database_info` reports them.

Over time

GitHub stars 105 +7 (+7 %)

10598

9 Oct 20267 Nov 2026

npm downloads a week 16k +15k (+4224 %)

16k49

6 Apr 20267 Nov 2026

Before the site measured it, read from npm's daily download counts.

Popularity score 62 +9 (+16 %)

6250

9 Oct 20267 Nov 2026

What is rising across the directory

Sources

Page checked 2026-10-15.

Badge for your README

Shows this server's rank in its category, and stars for a server that runs on your machine. It updates every hour.

MarkdownREADME.md

[![Dynoxide on EveryMCPServer](https://everymcpserver.com/badge/dynoxide.svg)](https://everymcpserver.com/servers/dynoxide)

Connect your client

Add in one click

Or add it by command or config

Add in one click

Or add it by command or config

Settings 1

What it does

What it needs

Limits

Who it suits

Tools 35

Similar servers

SocratiCode

dbt MCP Server

Instant MCP

autocontext

OpenComputerUse

Mobile MCP