Get account information (equity, cash, buying power) from a broker.

Get all open positions from a broker.

Get orders from a broker, optionally filtered by status.

Get a unified portfolio summary across ALL connected brokers — total equity, positions, and P&L.

Get current quote (bid/ask/last) for a symbol from a broker.

Return a unified health snapshot for all four live futures bots (MNQ, CL, MES, NQ) and the Kalshi daemon. For each bot: process up? last log mtime, last signal ts, current regime, error count in last 100 log lines, token expiry (if Tradovate). Includes E2E sentinel lifecycle state for MNQ execution traceability. Pure read-only — reads logs/, state/, and ps aux on the control tower host.

Hybrid (semantic + keyword + graph-traversal) search over the AlgoChains TEMPORAL knowledge graph (getzep/graphiti). Returns advisory facts with validity windows (valid_from/valid_to) extracted from REAL signal traces, debate transcripts, and Hive Brain synthesis. Use for 'what was true / what changed / what preceded what, over time' — e.g. 'MNQ behavior in trending regime'. agent_memory authority: ADVISORY ONLY, never broker truth (P&L/fills still require broker verification). Complements rag_search/onyx (semantic) and query_codegraph (structural). Fails closed with graphiti_unavailable.

Health probe for the Graphiti temporal knowledge-graph backend (Neo4j + graphiti-core, advisory/agent_memory). Reports provider, Neo4j URI, group_id, and reachability. Fails closed with graphiti_unavailable + recovery_command (per-host; not synced across machines).

Connect to a specific broker. Must be configured via environment variables.

Run the marketplace validation gates against reported strategy metrics (Sharpe, OOS trades, drawdown, win rate, MCPT). This is distinct from validate_strategy, which validates a StrategySpec schema.

Validate a StrategySpec for schema correctness, parameter ranges, and internal consistency.

Run a backtest on a StrategySpec using the Rust engine. Returns Sharpe, drawdown, win rate, P&L.

Run Optuna-based parameter optimization on a StrategySpec. Finds best params across n_trials.

BM25 search over all Massive market data API endpoints. Use this FIRST to find the right endpoint for stocks, options, futures, forex, crypto, or SEC filings.

Get parameter documentation for a Massive API endpoint. Pass the docs_url from massive_search_endpoints results.

Execute a Massive market data API call. Optionally store results as an in-memory DataFrame for SQL querying. Supports pagination auto-detection — check _next_page in results.

SQL queries over stored DataFrames from massive_call_api. Supports SHOW TABLES, DESCRIBE , DROP TABLE , and full SQL with JOIN/GROUP BY/window functions. Use apply for server-side Greeks and technicals.

Composable pipeline: search→fetch→store→query→apply in 1 call (saves 4 round-trips). Describe what data you want, optionally filter with SQL and apply Greeks/technicals.

Search for relevant AlgoChains tools using natural language. Returns the top-K most relevant tools with descriptions. Use this FIRST to find which tools are available for your task — 90%+ context reduction vs listing all 533 tools.

Get full details for a specific tool including its input schema, parameter types, and usage examples. Call after discover_tools to get the full spec before execution.

Execute any discovered tool by name with arguments. Use discover_tools first, then get_tool_details for the schema, then call this to execute. ORDER_EXEC and DESTRUCTIVE tools require owner_token and confirm=true inside arguments.

Return JSON manifest of all registered MCP tools with implementation_status (full|partial|stub), required env vars, and Tier-1 flags. Use for CI, Onyx indexing, and honest agent planning — call before relying on V8-V20 tools.

Transform a natural language trading intent into a concrete plan and execute it. Example: 'Get me $10K AI exposure, max 2% per stock'. Parses intent → solves constraints → presents plan for approval → executes.

Approve a pending intent plan for execution. The plan must be in 'pending_approval' status.

Create a shadow (paper) portfolio to forward-test a strategy without risking capital. Track P&L, fills, and metrics alongside your real portfolio.

Detect current market regime from VIX, SPY trend, breadth, and credit signals. Returns regime classification (bull/bear/range/volatile/crisis), recommended strategies, and risk multiplier for position sizing.

Run 13 pre-trade safety checks before placing an order. Checks position sizing, daily loss limits, drawdown, fat fingers, buying power, concentration, VIX killswitch, margin, correlation, and more. Returns ALLOW or BLOCK with reasons.

View current account protection settings including daily loss limits, drawdown thresholds, position size caps, VIX killswitch levels, and max positions.

Query AlgoChains data warehouses (Builder tier $199/mo). Access 3.09B+ rows: 409M crypto, 1.3B stocks, 1.4B forex minute bars. Returns OHLCV data for backtesting.

Submit a validated strategy to the AlgoChains marketplace. Runs 7-gate validation (schema, performance, overfitting, MCPT, walk-forward, paper trading, decay monitor). Returns tier classification (Platinum/Gold/Silver/Bronze) and next steps.

Compute full implied volatility surface from real Polygon options chain: IV per strike/expiry, 25-delta skew, term structure, IV rank (0-1), IV percentile, and vol regime (low/normal/elevated/extreme). Generates actionable signal: long_vol/short_vol/sell_skew/buy_skew.

Decompose a symbol's returns into Fama-French 5-factor + momentum exposures using real Polygon daily data. Returns alpha, market beta, SMB/HML/momentum betas, R-squared, information ratio, tracking error. Identifies alpha-generating vs factor-exposed regimes.

Detect market regime using Hidden Markov Model on real daily returns: bull_trending, bear_trending, choppy, or crisis. Returns regime probability, days in current regime, transition probabilities, vol regime, and Sharpe. Uses hmmlearn if available, statistical fallback otherwise. Real Polygon data only.

Aggregate shadow-only quant regime telemetry from bot_metrics_live and state/quant_shadow_snapshot.json: GARCH status, OFI intensity, Kalman shadow slope, HMM regime status, and 7-day agreement summary when available. Does not compute models.

Get VIX term structure from real CBOE data: spot VIX, VIX3M, VIX6M contango/backwardation. High contango (>10%) is bullish for equities; backwardation signals fear. Returns regime: contango, backwardation, flat.

Compute real-time cross-asset correlation matrix for a list of symbols using actual daily returns. Detects regime changes (correlation spikes during crises). Returns heatmap data, average pairwise correlation, and risk concentration score.

MCP Elicitation: request structured human confirmation before executing a high-value or destructive trade action. Shows the user a form with trade details; execution is gated on approval.

Submit a durable long-running MCP Task (backtest, optimization, ML retrain). Returns a task_id immediately. Use get_task_status to poll. Tasks persist across disconnects.

Get status and progress of a long-running MCP Task. Returns phase, progress percentage, result (when done), or error.

Trigger an AlphaLoop evolution cycle: SCAN underperformers → MUTATE parameters via Optuna → VALIDATE against real trade history → PROMOTE winner. Uses RL reward model. Requires real trade history (min 5 trades).

Compute footprint chart for a symbol: bid/ask volume at each price level per candle, detecting absorption (sellers absorbed at support), imbalance (>3:1 ratio), and delta exhaustion. Uses real Databento tick data.

Fetch dark pool volume for a symbol from real FINRA ATS reports + Polygon off-exchange trade conditions. Returns dark pool %, total off-exchange volume, and institutional activity score. NO synthetic data — fails if real sources unavailable.

Run earnings NLP pipeline: fetch SEC EDGAR filing, compute FinBERT sentiment, extract key themes (guidance, EPS beat/miss, capex), detect tone shift vs prior quarter. Returns catalyst score and actionable signal.

Fetch real prediction market probabilities from Polymarket and Kalshi for macro events (Fed rate decisions, election outcomes, economic releases). Derives equity market signals from contract odds.

Search live Polymarket and/or Kalshi markets by keyword. Returns real contract YES/NO prices, volume, liquidity, and URLs. Fails closed if no API data.

List highest 24h-volume Polymarket markets right now (real Gamma API). Useful for Roo-style early YES/NO flow and liquidity discovery.

Append one real performance snapshot for a Polymarket or Kalshi bot to the JSONL audit log (latency, YES prob, edge). Required for marketplace promotion evidence trail. No synthetic values stored unless caller passes them.

Read recent JSONL metric entries for a prediction-market bot_id from the local audit log.

Fetch detailed info for a specific Polymarket market by condition ID or event slug. Returns question, YES/NO prices, volume, liquidity, resolution date, and status. More precise than search — use when you have a specific market ID.

Get historical YES price data for a specific Polymarket market. Returns timestamped price series. Accepts slug, Gamma numeric ID, or CLOB token ID — auto-resolves. Useful for charting probability movement, analyzing market efficiency, and detecting smart money flow timing.

List Polymarket prediction markets with status filtering and pagination. Unlike search, this returns all markets in a category. status=open (default) | closed | resolved. Sorts by 24h volume descending.

Fetch recently settled Kalshi prediction market contracts (RSA-PSS signed). Returns results, profit-per-contract, and settlement timestamps. Requires KALSHI_ACCESS_KEY + KALSHI_PRIVATE_KEY_PATH. Inspired by 9crusher/mcp-server-kalshi settlements endpoint.

Read AlgoChains business identity files (TELOS system, adapted from PAI). Returns mission, goals, strategies, mental models, lessons learned, challenges, ideas, and KPIs. Use section='all' for full context or specify: mission|goals|strategies|models|learned|challenges|ideas|metrics. Every agent should read TELOS at session start for full business context.

Append a new entry to an AlgoChains TELOS file (goals, learned, ideas, challenges, etc.). Use to capture new lessons learned, ideas, or goal updates during a session. The log is append-only — entries are never overwritten.

Fetch US economic indicators from FRED (Federal Reserve Economic Data). Covers 16 key indicators: VIX, Fed Funds Rate, CPI, PCE, 10Y-2Y Treasury spread, unemployment, M2, GDP, housing starts, consumer sentiment. Requires FRED_API_KEY (free at fred.stlouisfed.org). Results cached 6h. Essential for regime detection across all bots.

Fetch EIA weekly crude oil inventory data — critical signal for the CL (crude oil) futures bot. Covers US commercial crude stocks, Cushing Oklahoma (WTI delivery point), and field production. Released every Wednesday ~10:30 AM ET. Build above estimate = bearish CL; draw below = bullish. Requires EIA_API_KEY (free at eia.gov/opendata).

Get the 7 most important Fed policy indicators in one call: Fed Funds Rate, CPI, PCE, 10Y-2Y spread, VIX, 10Y yield, 2Y yield — with AI-derived regime interpretation (restrictive/neutral/accommodative, crisis/normal, inverted/normal yield curve). Use for MNQ/NQ regime context before trading sessions. Requires FRED_API_KEY.

Record the outcome of an agent action or skill invocation for continuous learning. After 30+ signals, patterns emerge: which skills produce the best outcomes, where failure is common, what to improve. Stored in state/learning_signals.jsonl (append-only audit log). Use after any significant agent action.

Retrieve and analyze historical learning signals from state/learning_signals.jsonl. Returns signals with optional summary statistics: success rate by action type, top skills by effectiveness, bot activity, average ratings. Use to identify where agent performance is strongest/weakest and drive improvement priorities.

Send a mobile push notification via ntfy (https://ntfy.sh). Topics: bots (bot up/down/trade), risk (circuit breaker, daily loss), marketplace (new subscriber, bot promoted), ops (deploy, system health), alpha (high-confidence signal). Priority: max/urgent = always-on screen; high = with sound; default = normal; low/min = silent. Requires NTFY_BASE_URL + optional NTFY_AUTH_TOKEN.

Check if the AlgoChains Django signal propagation service (Roo architecture) is reachable and whether copy-trade paper fanout has active backlog. Separates active_lag_seconds from idle_since_last_signal_seconds so quiet markets do not look stalled.

Run the GUARDRAIL pre-flight middleware chain before placing any order. Executes 6 gates: VIX, daily-loss, stoploss-guard, cooldown, confidence, R/R. Returns approved=true only if all gates pass. Wire this before every order execution.

Get pre-computed macro alpha signal fabric: yield curve shape (2y-10y), credit spreads (HY-IG), DXY momentum, PMI regime, VIX term structure contango/backwardation. All from real FRED/CBOE/Polygon APIs.

Get real-time dashboard of all live trading bots: PIDs, positions, today's P&L, signal counts, win rates computed from actual fill history. Data from ~/.algochains/bot_metrics.db.

Subscribe to real-time bot metrics stream via MCP resource notifications. Fires on every fill, signal, and position update. Perfect for the private bot showcase on AlgoChains marketplace.

List all available AlgoChains skills from OpenClaw (363+), Windsurf (80+), Cursor (15), and Claude (8) skill libraries. Filter by category (trading, research, operations, intelligence, agent, comms, risk, data, ml, marketplace) or platform. Returns name, description, categories, tools used, and trigger type.

Get the full SKILL.md content and metadata for any skill by name (e.g. 'moltbook-debate', 'bot-diagnostics', 'autonomous-researcher', 'backtest-governance'). Returns complete instructions, tool requirements, trigger conditions, and schedule. Use list_skills or search_skills to discover skill names.

Search across all 450+ skills by keyword. Returns ranked matches from OpenClaw, Windsurf, Cursor, and Claude libraries. Use to find the right skill for a task before reading its full SKILL.md.

Given a task description in plain language, return the 3-5 best skills to use. Matches your task against skill descriptions across all platforms. Use when you do not know which skill to call.

Read the OpenClaw agent memory store. Contains trade lessons, regime history, signal quality scores, and cross-session agent context. Filter by key_prefix (e.g. 'trade', 'regime', 'bot') to narrow results.

Persist a trade lesson to OpenClaw memory so autonomous agents can learn from it. Lessons are retrieved during future trade decisions for similar setups. Required: symbol, direction, outcome, lesson text.

Read the current market regime from OpenClaw state (written by autonomous regime_detector skill). Returns regime label, confidence, and timestamp. This is the regime all live bots use for signal filtering.

Read ~/.openclaw/bot_heartbeat.json. This file is MNQ-only and fill-triggered (written by FUTURES_SCALPER_UPGRADED._track_openclaw_feedback on slippage/fill feedback), NOT by autonomous_watchdog every 5 minutes. Schema is typically {ts, bot, symbol}. For fleet process liveness use get_bot_health / get_all_bot_ops_status; for failover primary use control-tower logs/bot_heartbeat.json.

Get existence, size, and last-modified time for all OpenClaw state files (memory, regime, heartbeat, monitor, evaluations, AI cost, calibration). Use to verify OpenClaw is healthy and its state files are current.

Trigger a Moltbook bull/bear multi-agent debate for a trading signal. Shadow mode — does NOT place orders. Returns consensus direction, confidence, agreement %, and per-agent reasoning. Use before significant trades for multi-agent validation.

Run the MCPT marketplace autopilot pipeline. Steps: decay (check edge decay), graduate (30-day paper trading gates), audit (batch MCPT re-validation), listing (generate marketplace JSON), slack (post summary to #quant-lab). Calls scripts/mcpt_autopilot.py.

Run the regime detection pipeline — analyzes VIX term structure, market breadth, and price action to classify current market as trending/choppy/volatile/mean_reverting. Updates OpenClaw current_regime.json used by all live bots.

Semantic search over the AlgoChains Onyx knowledge base: 400+ strategy research JSONs, 45+ blueprints, 126 skills, live bot logs. Returns ranked documents with relevance scores.

Ask a natural language question against the Onyx knowledge base with RAG grounding. Returns an answer with cited sources. E.g. 'What is the best CL swing setup in trending regimes?' or 'How do I configure Token Guardian?'

Get real-time perpetual futures funding rates from Binance, Bybit, and Hyperliquid. Identifies funding rate arbitrage opportunities and predicts funding-driven price pressure.

Get real staking APY from Lido Finance (stETH), Binance Simple Earn, Cosmos validators, and Ethereum Beacon Chain. Compares yield opportunities across protocols.

Get status and result of a dispatched tower job. Polls the tower via SSH for the result file.

Check the configured compute node (ALGOCHAINS_TOWER_HOST) health: reachable, memory, active jobs, GPU status.

Run the autonomous marketplace pipeline: Research→Backtest→MCPT Validate→Stage for marketplace. Scans recent strategy research, runs tick backtests, applies 5-gate validation, stages passing strategies as marketplace JSON listings. Triggers Onyx ingest and Slack notification. No synthetic data — real tick engines only.

Get all staged marketplace bot listings with real metrics: futures (owner-only), equities, crypto, forex. Includes Sharpe, win rate, max DD, subscription pricing, and paper trading status. Supabase-first with local filesystem fallback.

Trigger an incremental Onyx knowledge base ingest: indexes new strategy research, marketplace listings, blueprints, skills, and bot logs into the self-hosted Onyx RAG host (ONYX_API_URL).

Check Onyx knowledge base status: health, last sync time, total indexed documents, connector status (self-hosted host via ONYX_API_URL).

Check AlgoChains Learn Hub health: HTTP status of /learn/, /learn/feed.xml RSS MIME, and learn.algochains.ai subdomain redirect. Read-only — does NOT deploy. Use to verify the live Learn Hub is up and public (no login required).

Get real-time trading metrics for live bots (Tradovate + Alpaca paper). Supabase-first (bot_metrics_live table). Returns daily P&L, win rate, last signal, confidence, error count. Bot IDs: mnq, cl, mes, nq, alpaca_paper_equities, alpaca_paper_crypto. Omit bot_id to get all. Falls back to log parser if Supabase unavailable.

Get real-time trading metrics for all 4 live Tradovate bots (MNQ, CL, MES, NQ) in a single call. Returns daily P&L, win rates, signals, error states, and MCPT validation badges. Data from real log files.

Check whether this MCP server node is the primary trader (MacBook offline) or standby (MacBook alive). Reads the Mac heartbeat file to determine heartbeat age, Mac liveness, desktop bot process counts (expected 5: MNQ/CL/MES/NQ + Kalshi), and which node is currently running the bots. Critical for dual-node failover awareness.

Read adaptive_brain.py daemon liveness from bounded process, script, state, and log evidence. Read-only; does not restart or mutate daemon state.

Run the trading-system-health audit: bot process/log liveness (with legacy log alias resolution), disk space on control-tower and home volumes, and optional health_snapshot.json. Use to triage SEV1 trading-system-health watchdog alerts without false inactive signals from stale cl_bot_live.log.

Get all academic citations, SSRN papers, and published works that provide the theoretical basis for a specific bot's strategy. Includes authors, year, venue, DOI/SSRN link, and relevance explanation. Bot IDs: mnq, cl, mes, nq.

Get the complete bot card data payload for algochains.ai marketplace display. Includes strategy summary, academic citations, backtest artifact paths (MCPT JSON, whitepapers, blueprints), skills references, and subscription tier. Use to populate or refresh a bot card on the marketplace site.

List all research attachments available for a bot: MCPT validation JSON files, backtest PDFs, whitepapers, and blueprint markdown files. Shows local path and whether the file exists. Use to prepare uploads to Supabase storage for bot card attachment panel.

Read the persisted position state file for a bot. Returns direction (BUY/SELL/null), qty, entry_price, and flat status. This is the bot's internal tracking — compare to Tradovate get_positions() to detect drift.

Parse the bot log to determine current bracket order status. Returns mode (live/oso_only/none/unknown), stop/target order IDs and prices, and whether the position is unprotected. Critical for detecting missing stops after an entry.

Check AI ensemble/debate pipeline health. Detects Anthropic quota errors, Cerebras model errors (llama3.1-8b), pipeline timeout events, and shadow mode. The pipeline is ADVISORY ONLY — primary confidence gate controls all trades regardless of pipeline state.

Full operational snapshot for all 4 bots: process status, PIDs, position states, bracket status, and AI pipeline health. Use to triage any bot integrity issue in one call.