Skip to main content
Glama
alforge-labs

alpha-forge-mcp

Official
by alforge-labs
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Local

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
ALPHA_FORGE_BINNoPath to the alpha-forge binary. If not set, the binary is searched on PATH or in default install locations.

Instructions

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

This server publishes no instructions, or was last inspected before Glama recorded them.

Capabilities

Features and capabilities supported by this server

Protocol revision2025-11-25

CapabilityDetails
tools
{
  "listChanged": false
}
prompts
{
  "listChanged": false
}
resources
{
  "subscribe": false,
  "listChanged": false
}
experimental
{}

Tools

Functions exposed to the LLM to take actions

NameDescription
list_strategiesA

List all registered AlphaForge strategies (strategy_id, name, version, timeframe).

get_strategyA

Get the full JSON definition of a registered strategy by its strategy_id.

list_resultsA

List saved backtest results, optionally filtered by strategy_id.

get_resultA

Get metrics and trades for a saved backtest result (result_id = strategy_id or run_id).

run_backtestA

Run a backtest for symbol with a registered strategy. Optional dates are YYYY-MM-DD.

Prerequisite: call `fetch_data` for the symbol first so the OHLCV cache exists.
Long-running: up to a 300-second timeout; reports progress to capable clients.
run_optimizeA

Optimize strategy parameters with Optuna for symbol. metric defaults to sharpe_ratio.

save defaults to true so the result JSON is persisted (with `saved_path` in the
response) and can be fed to `apply_optimization`; pass save=false to skip saving.
Long-running: up to a 600-second timeout; reports progress to capable clients.
generate_pinescriptB

Generate TradingView Pine Script v6 for a strategy. Returns {strategy_id, pinescript}.

run_walk_forwardA

Run walk-forward optimization for symbol (out-of-sample robustness check).

windows defaults to 5, metric to sharpe_ratio. Run it after run_optimize to compare
in-sample vs out-of-sample behaviour (the optimize_and_verify workflow).
Long-running: up to a 600-second timeout; reports progress to capable clients.
run_monte_carloA

Run a Monte Carlo simulation from a saved backtest result (resamples its trades).

Prerequisite: a saved result (run_backtest/run_optimize with save) — result_id =
strategy_id or run_id. simulations defaults to 1000. Returns ruin probability, equity
percentiles, and drawdown distribution for risk assessment.
Long-running: reports progress to capable clients; has an execution timeout.
fetch_dataA

Fetch & cache historical OHLCV for symbol (prerequisite for run_backtest).

period is e.g. 1y / 5y / 6m / 30d / max (defaults to 1y). Returns {symbol, period,
output}. The CLI has no --start/--end, so only period is exposed. Run this before
run_backtest. Reports progress to capable clients; has an execution timeout.
save_strategyA

Register a strategy from its JSON body (not a file path; agent-friendly).

Pass the full strategy-definition JSON as a string; it is validated as a JSON object
and written to a temp file before `strategy save`. Returns {output}. A registered
strategy is the prerequisite for run_backtest/run_optimize. Reports progress to
capable clients; has an execution timeout.
forge_statusA

Report alpha-forge capabilities/prerequisites before use (doctor + version).

Read-only triage: returns {binary_found, version, authenticated, plan, doctor, error}.
Never fails when the binary is missing — returns binary_found=false instead.
apply_optimizationA

Apply an optimization result file to a strategy, saving <strategy_id>_optimized.

Prerequisite: run_optimize(save=true) — result_file is its `saved_path`. Runs
non-interactively (--yes). Returns {result_file, strategy_id, output}. Follow up by
generating Pine Script for `<strategy_id>_optimized`. Reports progress to capable
clients; has an execution timeout.
list_journalsA

List strategies that have a journal (history of snapshots and runs).

get_journalA

Get the full journal (snapshots, runs, tags, notes) for a strategy_id.

exploration_statusA

Show the strategy-exploration coverage map (explored vs. untried combos).

Optional `goal` filters by exploration goal; defaults to the "default" goal.
get_indicatorA

Get metadata for a technical indicator (description, parameters, output, example).

`indicator` is the indicator name (e.g. RSI, MACD). This is metadata only — the CLI
has no compute-over-symbol command — so it does not run a calculation on price data.

Prompts

Interactive templates invoked by user choice

NameDescription
backtest_and_reviewGuide: run a backtest for a strategy/symbol, then review the metrics.
optimize_and_verifyGuide: optimize a strategy with Optuna, then check it is not overfit.

Resources

Contextual data attached and managed by the client

NameDescription
resource_strategiesAll registered strategies (same payload as the list_strategies tool).
resource_resultsAll saved backtest results (same payload as the list_results tool).
Deploy Server
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

A3.9/5.0

Scored across 17 tools

Disambiguation4/5

Most tools have distinct purposes. run_optimize and run_walk_forward both optimize but are differentiated by description (single vs. walk-forward). get_result, get_strategy, and get_journal serve different retrieval needs. Minor overlap but clear descriptions guide selection.

Naming Consistency4/5

Tools predominantly follow verb_noun pattern with underscores (e.g., run_backtest, get_strategy). One tool, exploration_status, uses noun_noun, breaking the pattern slightly. Overall consistent and readable.

Tool Count5/5

17 tools are well-scoped for a quantitative backtesting platform, covering data fetching, strategy management, backtesting, optimization, Monte Carlo, walk-forward, and results. Each tool has a clear role without bloat.

Completeness4/5

The tool surface covers the full lifecycle from data fetch to strategy registration, backtesting, optimization, and result analysis. Minor gaps like missing delete/update tools for strategies or results, but core workflows are intact.

Maintenance

ActivitySlowing
ResponsivenessUnresponsive

Resources