list_all_challenges
List all active CrowdCent challenges.
LABSCO SUMMARY
The whole competition loop fits in one conversation.
A coding agent can pull the data, train in your workspace and submit the predictions in a single session; a chat-only client fetches through the signed-URL tools and submits inline from a dataframe. The hosted server stores nothing — your key authenticates each request exactly as the Python client does.
What it is
The CrowdCent MCP server, which now ships inside the crowdcent-challenge package rather than from this repository. It downloads challenge data, submits predictions, reads back how they scored, and backtests portfolio constructions on the community meta-model.
What you get
- Signed download URLs from get_training_dataset_url, get_inference_data_url and get_meta_model_url, usable by any agent with a terminal or a fetch tool
- Prediction submission, plus your recent submissions and how each one scored
- Simulation over the meta-model with the same engine and tier gates as the site's Simulation tab — sweep configurations, blend sleeves, compare in-sample against out-of-sample
- The local server additionally carries the file-download and file-submission tools
- switch_challenge is gone: challenge-scoped tools now take an optional challenge_slug, defaulting to hyperliquid-ranking
Requirements
Two paths, both replacing a local clone of this repository. Hosted: add https://mcp.crowdcent.com/mcp to any client that supports remote MCP over HTTP, with your API key as a bearer token — it is a stateless pass-through that stores nothing. Local: one uvx line, uvx --from crowdcent-challenge[mcp] crowdcent-mcp, with CROWDCENT_API_KEY in the env block; no clone and no absolute paths. Keys come from your crowdcent.com profile settings. If a config still points at a local server.py, replacing that one entry is the whole migration.
Setup effort
One command plus a key — claude mcp add --transport http crowdcent https://mcp.crowdcent.com/mcp --header "Authorization: Bearer <YOUR_API_KEY>"
USE CASES
01
Getting the training and inference data
Signed URLs the agent can fetch directly, rather than files written to disk it then has to find.
02
Backtesting the community signal
Sweep leg size and cadence over the meta-model and see where the stable region is, in-sample and out.
03
Submitting predictions and seeing how they scored
Submit for the current inference period, then read the recent submissions and their scores back.
04
Moving off a local clone
One hosted URL or one uvx line replaces the directory-and-path config, and the server always matches the client library it ships with.
30 tools
Listed in project docs. Your client may show a different list.
get_challenge_info
Get detailed information about a challenge.
list_training_datasets
List all training dataset versions for a challenge.
get_training_dataset_info
Get details for one training dataset version ('latest' works).
get_inference_data_info
Get details for one inference period ('current' works).
get_training_dataset_url
Get a signed, time-limited download URL for a training dataset.
get_inference_data_url
Get a signed, time-limited download URL for inference features.
get_meta_model_url
Get a signed, time-limited download URL for the consolidated meta-model file.
submit_predictions_from_dataframe
Submit predictions from a JSON representation of a dataframe.
list_submissions
List your prediction submissions for a challenge.
get_submission
Get details for one of your submissions by id.
get_performance
Get your historical submission performance (scores flattened per row, newest first).
download_training_dataset
Download a training dataset to a local parquet file.
download_inference_data
Download inference features to a local parquet file.
download_meta_model
Download the consolidated meta-model to a local parquet file.
submit_predictions_from_file
Submit predictions from a local parquet file.
run_simulation
Backtest one portfolio config against the live meta-model.
sweep_simulations
Grid-search up to your tier's budget (96 configs at Contender, 24 below) of portfolio constructions in one call.
blend_simulations
Blend up to 3-5 weighted sleeves (tier-capped) into one ensemble book; returns blend stats plus the sleeve correlation matrix. Sizing is the blend's, never a sleeve's: weights shape the blend, and the netted book is sized once by leverage / target_vol.
get_trading_accounts
List the user's Hyperliquid trading accounts (both networks): status, can_trade, master_address, agent expiry, builder approval. Never any key material.
get_mandate
Get the live mandate (execution policy + weighted strategy sleeves) for this challenge on the given network.
set_mandate
Create or fully replace the mandate: {"sleeves": [{"config" | "config_token", "weight", "label"?}], ...execution knobs}.
get_target_book
The blended target book the mandate currently resolves to: natural-gross target holdings, ranking day, per-sleeve books, and the one gross_multiplier the planner deploys (the mandate's leverage, or the smaller multiple its vol target picked under it).
preview_rebalance
Plan a rebalance WITHOUT executing: trades, turnover, estimated fees, plus the plan_hash needed for execution. Always the first step — show the user this plan before anything live.
execute_rebalance
Execute a LIVE rebalance on the user's Hyperliquid account.
flatten_positions
Close EVERY position (full liquidation) — two-step like rebalancing: call with preview=true first, show the user the plan, then call again with the returned plan_hash after their explicit confirmation.
pause_trading
Pause the mandate: scheduled trading stops until resumed. Safe to call proactively when something looks wrong.
resume_trading
Resume the mandate (scheduled trading on). Confirm with the user before re-enabling live risk.
list_rebalance_runs
Recent rebalance runs (the audit trail): kind, status, ranking day, result summary, error. The morning-briefing feed.
list_open_orders
The order blotter, newest first, optionally filtered by status (resting, filled, twap_running, ...).
BEFORE YOU INSTALL
What to know first
Tools you get
30 tools named in the project's documentation.
Requires
- A client that can reach a remote MCP server over HTTP
- CROWDCENT_API_KEY in your environment
Where it runs
Of the 4 documented clients, 1 runs the server on your machine and 3 connect to a hosted endpoint.
INSTALLATION
Connect your client
Server setup
claude mcp add
Claude Code — hosted
TERMINAL
claude mcp add --transport http crowdcent https://mcp.crowdcent.com/mcp --header "Authorization: Bearer <YOUR_API_KEY>"Hosted endpoint
ENDPOINT URL
https://mcp.crowdcent.com/mcpAdd this hosted endpoint to your MCP client using the credentials listed under Requirements.
MCP Clients
Cursor — hosted
CURSOR — HOSTED · JSON
{
"mcpServers": {
"crowdcent": {
"url": "https://mcp.crowdcent.com/mcp",
"headers": {
"Authorization": "Bearer <YOUR_API_KEY>"
}
}
}
}VS Code — hosted
VS CODE — HOSTED · JSON
{
"servers": {
"crowdcent": {
"url": "https://mcp.crowdcent.com/mcp",
"headers": {
"Authorization": "Bearer <YOUR_API_KEY>"
}
}
}
}Any MCP client — uvx
ANY MCP CLIENT — UVX · JSON
{
"mcpServers": {
"crowdcent": {
"command": "uvx",
"args": [
"--from",
"crowdcent-challenge[mcp]",
"crowdcent-mcp"
],
"env": {
"CROWDCENT_API_KEY": "<YOUR_CROWDCENT_API_KEY>"
}
}
}
}
