DynoTable - DynamoDB GUI with Real SQL and AI Agent

TL;DR

TL;DR: DynoTable is a local-first DynamoDB GUI that runs real SQL (JOINs, GROUP BY, aggregates) against your tables, includes an AI agent powered by your own AWS Bedrock keys, and exposes an MCP server for Claude Code and Cursor. Free read-only plan, 30-day full trial.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: dynotable.com ← MUST visit and verify
• HN launch thread: news.ycombinator.com/item?id=49082126 ← verified via HN Algolia API
• Source (GitHub): github.com/dynotable ← verified via curl github.com/dynotable/dynotable

What Is DynoTable?

DynoTable is a desktop DynamoDB client that addresses three long-standing pain points with official and third-party GUI tools:

1. Real SQL over DynamoDB — Most GUIs give you only PartiQL (DynamoDB’s SQL-like subset) or a visual query builder. DynoTable compiles real SQL — including JOIN, GROUP BY, COUNT, SUM, AVG, and other aggregates — into DynamoDB’s own Query and Scan operations. It stays within your table’s access pattern rules while giving you the expressiveness of SQL.

2. AI agent on your own keys — An AI assistant reads your schema, writes queries, and stages edits for your review. It runs entirely on your own AWS Bedrock account — prompts, schema, and row data never leave your AWS environment. No per-seat AI surcharge; you pay AWS at cost.

3. MCP server for Claude Code and Cursor — DynoTable runs a local MCP server. Connect Claude Code or Cursor directly to your tables, with schema introspection and item reads on the free tier. Paid plans add SQL workbench and write tools.

Setup Workflow

Step 1: Download and Install

DynoTable runs as a desktop app on macOS, Windows, and Linux.

macOS:

# Download from dynotable.com/download
# Or install via Homebrew (if available)

Windows: Download the x64 installer from dynotable.com/download.

Linux: Available as .deb and .rpm packages from the same download page.

Step 2: Connect Your AWS Credentials

On first launch, add an AWS profile. DynoTable supports:

• Standard AWS credentials (access key + secret)
• AWS IAM Identity Center SSO — sign in with the standard device flow directly in the app, no AWS CLI required

Your credentials are stored locally and never sent to DynoTable’s servers.

Step 3: Browse and Query

Once connected, you see a visual table browser withPartiQL support and a SQL Workbench.

Example: Run a real SQL join across two tables

SELECT orders.customer_id, customers.name, SUM(orders.amount) AS total
FROM orders
JOIN customers ON orders.customer_id = customers.id
GROUP BY orders.customer_id, customers.name
ORDER BY total DESC
LIMIT 20

DynoTable compiles this to DynamoDB Query/Scan calls, reads only the rows touched, and runs the join locally — no scan of the entire table.

Step 4: Connect Claude Code via MCP

DynoTable exposes a local MCP server. To connect:

1. Open DynoTable → Settings → MCP
2. Enable the MCP server (off by default)
3. In your Claude Code project, add the MCP server connection
4. Grant per-connection consent — you choose exactly what the agent can access

# Example MCP server endpoint (shown in DynoTable MCP settings)
mcp://localhost:8765

The agent can read schema and items on the free tier. Paid plans unlock the SQL workbench and staged write tools.

Step 5: Stage and Commit Changes

Any edit you make — via the UI or through the AI agent — collects in a staged commits window before touching DynamoDB. Review all pending changes, then commit atomically. This prevents accidental table mutations.

Deeper Analysis

Local-First Architecture

DynoTable’s design philosophy is local-first. Your data and credentials never leave your machine. The SQL engine, AI inference (on Bedrock), and credential storage all run locally or against AWS directly. This is a meaningful distinction from cloud-based DynamoDB GUIs that route traffic through their own servers.

AI Agent Privacy

The AI agent uses your own AWS credentials to call Bedrock. From AWS’s perspective, the requests come directly from your account. DynoTable’s server is not in the call chain at any point.

Pricing for the AI is AWS Bedrock at cost — typically $0.001–0.005 per query depending on model and table size. There is no DynoTable markup.

Query Cost Estimation

Before executing a SQL query, DynoTable previews:

• Whether it will use Query (efficient) vs Scan (expensive)
• Which index will be used
• A read capacity unit (RCU) estimate

This prevents accidental full-table scans that could inflate your DynamoDB bill.

Staged Commits

Every write operation — whether triggered via the UI or the AI agent — lands in a staging window. You review the full diff before committing. This is similar to a git staging area, applied to DynamoDB mutations.

Free vs Paid

| Plan | Price | Features | |------|-------|----------| | Free | Forever | Browse, PartiQL queries, CSV/JSON/NDJSON export, MCP schema + item reads | | Individual | From $9/mo per seat | SQL Workbench, Smart Tables (saved joins), editing, staged commits, AI agent | | Team | Custom, up to 20 seats | Everything in Individual + shared workspace, SSO, seat management, shared indexes |

The 30-day trial requires no credit card. When it ends, the app stays on the Free plan — you are never locked out.

Practical Evaluation Checklist

• Does it run real SQL JOINs across DynamoDB tables?
• Does the AI agent stay on your own AWS account?
• Is there a free forever plan with read-only access?
• Can Claude Code or Cursor connect via MCP?
• Are writes staged before committing to DynamoDB?
• Does query preview show RCU estimates?

Security Notes

• AWS credentials stored locally; never transmitted to DynoTable servers
• AI inference via your own Bedrock account — data never touches third-party servers
• MCP server runs locally; external agents must be granted explicit per-connection consent
• IAM permissions enforced at the credential level — a read-only IAM role means read-only access in DynoTable

FAQ

Q: Does this work with DynamoDB Local? A: Yes. Set a profile’s endpoint to http://localhost:8000 (DynamoDB Local) and work entirely offline. Profiles can each target a different region or a local endpoint.

Q: Can I use Claude Code with my tables? A: Yes. DynoTable’s MCP server exposes schema introspection and item reads on the free tier. The AI agent and write tools require a paid plan.

Q: How is this different from Dynobase? A: DynoTable runs real SQL with JOINs, GROUP BY, and aggregates — Dynobase does not. DynoTable also has a local-first architecture and an AI agent that runs on your own Bedrock keys, not on Dynobase’s servers.

Q: Does the AI send my data to a third party? A: No. The agent calls AWS Bedrock directly using your credentials. There is no DynoTable server in the request path.

Q: What happens when the trial ends? A: You are moved to the Free plan (read-only). Browse, query with PartiQL, export, and MCP schema reads continue to work. SQL Workbench, editing, and AI agent require a paid plan.

Q: Which platforms are supported? A: macOS (Apple Silicon), Windows (x64), and Linux (.deb and .rpm) are all available.

Conclusion

DynoTable fills a specific gap in the DynamoDB tooling ecosystem: a GUI that speaks real SQL, keeps credentials and data local, and integrates AI assistance without a per-seat AI tax. The free-forever read-only plan makes it easy to evaluate — download, connect, and query without handing over a credit card.