
algorover-research
io.github.algorover/algorover-research
Discover trading strategies, run historical backtests, and inspect research reports.
Documentation
AlgoRover Research
English | 简体中文
AlgoRover Research helps you find published trading strategies, run historical backtests on public market data, and inspect trades and research reports through conversation. Backtests run in the hosted AlgoRover Research service.
This repository distributes version 0.1.5 of the Codex plugin and its installation package. The plugin connects to https://research.algorover.ai/mcp. You do not need to install Python, Docker, the main AlgoRover application, or a local Runtime.
Version 0.1.5 preserves the service-provided strategy name, version and task ID when saving report downloads, and adds a numeric suffix when a file with the same name already exists. It retains Chinese and English conversation-language handling and authorization recovery guidance.
Research cards show a release notice with update instructions. Check the installed version in Codex plugin details; the card cannot detect it automatically. You can dismiss the notice for the displayed version. Reopen the card after updating. The notice does not block backtests.
Features
- Search strategies and inspect published versions, package licenses, and runnable presets.
- Prepare a backtest using a supported market and preset, choose a 90-day or 365-day historical window, and review the configuration before starting.
- Run backtests, check progress, and cancel unfinished tasks.
- View your backtest metrics, trade details, and task history, and download research reports. Request download links for published strategy packages when needed.
- Delete a selected backtest result or all your Research data on request. Deletion requires separate authorization.
Available strategies, markets, and presets depend on what the service returns. Strategies that need preset preparation go through that step first. If preparation fails, the plugin displays the reason returned by the service.
Installation and sign-in
Use a Codex client that supports plugins. To install from a terminal, run these commands in order:
codex plugin marketplace add algorover/algorover-research
codex plugin add algorover-research-cloud@algorover-research-github
codex mcp login algorover-research --oauth-client-registration dcr --scopes research:read,research:run
The last command opens the authorization page in your browser. Sign in with your AlgoRover account, review the requested permissions, and allow access.
In Codex, select @AlgoRover Research and say “I want to start using AlgoRover Research.” The interactive strategy catalog opens in the chat. Choose a strategy on the card to see its details and prepare a backtest. If Research tools are unavailable, check the plugin connection and follow Connection issues below.
If you already installed the plugin from this repository, refresh the marketplace and reinstall the plugin from it:
codex plugin marketplace upgrade algorover-research-github
codex plugin add algorover-research-cloud@algorover-research-github
Strategy search and backtest configuration previews are publicly accessible. Starting backtests, viewing personal tasks, and downloading personal reports require sign-in. The command above requests these permissions:
| Permission | Purpose |
|---|---|
research:read | Read research tasks, results, and reports belonging to the signed-in account. |
research:run | Start and cancel backtest tasks belonging to the signed-in account. |
Deleting data requires the additional research:delete permission, which the command above does not request. To use deletion tools, authorize again through your client with that permission, or manage your data on the Research website.
Sign-in and authorization take place in your browser. Do not share passwords, access tokens, or exchange API keys in the conversation.
Example prompts
- "Use AlgoRover Research to find EMA pullback strategies and show their runnable presets."
- "Choose an available ETH preset and prepare a backtest for the last 90 days. Show me the configuration first."
- "Start the backtest with that configuration and check its progress."
- "Show the metrics and trades from my latest successful backtest, and give me the report download link."
- "Cancel the backtest I just started."
Data and permissions
The plugin sends the strategy reference, market, preset, and historical window needed for your task to the AlgoRover Research service. Tasks and results belong to the signed-in account and can be retrieved across conversations. Access to personal data follows account permissions.
You can request deletion of a selected backtest result or explicitly request deletion of all your Research data. Deleting all Research data is separate from deleting the identity account used to sign in. See the privacy policy for data collection, use, and retention rules.
AlgoRover Research supports historical research only. It does not connect exchange accounts or execute trading orders. Historical backtest results do not establish future returns.
Connection issues
If you see OAuth authorization required, reauthenticationRequired, or notLoggedIn, reconnect Research through the client or run the sign-in command above. Complete browser sign-in and approve the plugin's requested permissions, then retry your original request after the client confirms the connection. Google identity confirmation alone does not confirm Research sign-in or the plugin connection. If the callback times out, start a fresh authorization attempt. Start a new conversation only if the connected client requires it to load tools.
After deletion of all Research data completes, sign in and authorize again with the same AlgoRover account to start new research. Deleted tasks, results, and reports stay unavailable. Access and authorization remain paused while deletion is pending.
If the service cannot be reached, check your network and the client's MCP connection status. When contacting support, include the error message and request ID. Leave out passwords and tokens.
Publisher and support
Published by Rivendell Holdings Investment Limited.
No installation packages available.