A Customer Used Manus AI to Query Our Data — Here's What That Means

· 4 min read
ai-agents manus tutorial motherduck duckdb

A subscriber recently told us they used Manus AI to retrieve HKJC pre-race odds from their Renavon subscription. They didn't configure an MCP server. They didn't write Python. They described what they wanted — pre-race odds for the current racing season — and Manus handled the rest: authenticating, connecting to their database, writing SQL, and returning the data.

This is different from the MCP-based workflow we wrote about earlier this month. MCP connects AI assistants like Claude or Cursor to your data through a protocol — powerful, but it requires developer setup. Manus and similar autonomous agents skip that entirely. You give them credentials and a goal, and they figure out the implementation themselves.

What is Manus AI?

Manus is an autonomous AI agent, now part of Meta after a $2B+ acquisition in late 2025. The name comes from the Latin for "hand" — the idea being that it doesn't just advise, it acts.

Where ChatGPT tells you how to query a database, Manus queries the database. It runs in a cloud sandbox with access to a web browser, a Python runtime, and shell commands. You describe a task in plain language, it breaks it into steps, writes and executes code, and delivers the output — a spreadsheet, a report, a chart, whatever you asked for. It works asynchronously: assign the task, walk away, come back to results.

It's available via web at manus.im, plus Telegram, WhatsApp, Slack, and LINE. There's a free tier and paid plans starting at $19/month.

How It Works with Renavon Data

Renavon datasets live on MotherDuck — cloud-hosted DuckDB databases that any standard DuckDB client can query. The connection pattern is straightforward:

  1. Call POST /api/v1/session-token with your API key to get a short-lived MotherDuck token
  2. Connect via DuckDB using the token
  3. Query your databases

When you tell Manus something like "Using my Renavon API key, get all pre-race odds for HKJC races this season", it works through exactly these steps. It reads the API documentation (or figures it out from the endpoint responses), writes Python with the duckdb library, executes the query, and formats the results.

No special Renavon integration exists inside Manus. It works because our data is exposed through standard, well-documented interfaces — a REST API for authentication and DuckDB/SQL for queries. Any tool that can run Python can do the same thing.

Not Just Manus

The interesting part isn't Manus specifically — it's the pattern. Any autonomous AI agent that can execute code can query Renavon data the same way:

  • Manus (Meta) — cloud sandbox with Python + browser
  • ChatGPT with Code Interpreter — can run Python, connect to APIs
  • Devin (Cognition) — software engineering agent
  • Any future agent that can make HTTP requests and run SQL

The common thread: your data needs to be query-ready behind a standard API. That's why we built the platform on MotherDuck with a simple token-based authentication flow. We didn't anticipate Manus specifically, but we designed for exactly this kind of programmatic access.

For developers who want tighter integration — where your AI assistant has persistent, always-on access to your data — the MCP approach is still the better fit. But for ad-hoc analysis where you just want to say "get me this data" and have it happen, autonomous agents are hard to beat.

Tips for Best Results

If you're using Manus or a similar agent with Renavon data:

Be specific about what you want. "Get me HKJC race results" is vague. "Get all Group 1 race results from Sha Tin for the 2024/25 season with jockey, trainer, and finishing position" gives the agent enough to write a precise query.

Tell the agent your database names. Your databases are listed on the Connect page — names like renavon_hkjc_races or renavon_hkjc_race_results. Including these in your prompt saves the agent from guessing.

Provide your API key explicitly. The agent needs your Renavon API key to authenticate. You can find it on the Connect page. Tell the agent: "Use API key abc123 to call POST https://renavon.com/api/v1/session-token, then connect to MotherDuck with the returned token."

Check your subscription coverage. Each Renavon package gives access to specific datasets. If the agent reports that data is missing, it may be that the dataset isn't in your current subscription. Check the datasets page to see what's available.

Get Started

All you need is a Renavon subscription and an API key. Browse available datasets, grab your API key from the Connect page, and point your agent at POST https://renavon.com/api/v1/session-token. If you'd rather wire up persistent access through Claude, Cursor, or VS Code, follow the MCP setup guide instead.

No account yet? Sign up and subscribe — your data will be query-ready in minutes.

← Back to blog