AlgoVault x CrewAI – Composite Trade Calls for Your CrewAI Crew

90.47% PFE Win Rate · 96,864+ calls · 38+ Merkle-verified on-chain batches · 738 assets covered.

Don’t trust – verify the track record → Snapshot taken 2026-05-18 – live numbers refreshed in-page from https://algovault.com/api/performance-public

AlgoVault MCP gives your CrewAI agents a composite verdict in one call – direction, confidence, regime, and cross-venue funding/sentiment context – backed by a publicly auditable record anchored to Base L2 (agentId 44544). Drop it into any Crew or Agent workflow via crewai-tools and the official MCPServerAdapter.

Provenance: crewai==1.14.4 + crewai-tools[mcp]==1.14.4 on PyPI (verified 2026-05-18). Streamable-HTTP transport documented at https://docs.crewai.com/en/mcp/overview and https://docs.crewai.com/en/mcp/streamable-http. Adapter source: crewAIInc/crewAI-tools. AlgoVault MCP endpoint: https://api.algovault.com/mcp.

What you’ll build (90s read)

A runnable Python script that:

1. Connects to AlgoVault MCP at https://api.algovault.com/mcp via MCPServerAdapter (streamable-HTTP).
2. Loads the 4 AlgoVault tools as CrewAI BaseTool objects.
3. Invokes get_trade_call for any coin + timeframe and prints the parsed verdict.
4. (Optional) Wires the tools into a Crew with a research Agent so an LLM can decide when to call which tool.

Prerequisites

1. Python >=3.10, <3.14 (python3 --version). The crewai and crewai-tools packages require it.
2. AlgoVault skills plugin (optional helper pack for Claude Code / Cursor / Cline):

   claude plugin install AlgoVaultLabs/algovault-skills
   

3. Install the demo deps (pinned in examples/crewai/requirements.txt):

   pip install -r examples/crewai/requirements.txt
   

4. (Optional) LLM API key – only needed for the follow-on Crew + Agent snippet below. The bare python demo.py BTC 4h call works without one.

Demo: Direct trade-call read (<=80 lines)

# examples/crewai/demo.py (excerpt -- see file for full source)
import json
from crewai_tools import MCPServerAdapter

ALGOVAULT_MCP_URL = "https://api.algovault.com/mcp"


def fetch_trade_call(coin: str, timeframe: str) -> dict:
    server_params = {
        "url": ALGOVAULT_MCP_URL,
        "transport": "streamable-http",
    }
    with MCPServerAdapter(server_params) as tools:
        by_name = {t.name: t for t in tools}
        raw = by_name["get_trade_call"].run(
            coin=coin,
            timeframe=timeframe,
            includeReasoning=False,
            exchange="BINANCE",
        )
    payload = json.loads(raw)
    return {k: payload[k] for k in ("call", "confidence", "indicators", "regime")}


if __name__ == "__main__":
    result = fetch_trade_call("BTC", "4h")
    print(json.dumps(result, indent=2))

Run it:

python examples/crewai/demo.py BTC 4h

Sample output

{
  "call": "HOLD",
  "confidence": 13,
  "indicators": {
    "funding_rate": 0.00005632,
    "funding_24h_avg": 0.00005632,
    "funding_state": "ELEVATED",
    "oi_change_pct": 0,
    "volume_24h": 8144688081.84,
    "trend_persistence": "MEDIUM",
    "breakout_pending": "INACTIVE"
  },
  "regime": "TRENDING_DOWN"
}

The 4 keys (call, confidence, indicators, regime) are the documented contract. The full MCP response also includes price, reasoning, timestamp, _algovault provenance, and closest_tradeable – all available in payload if your agent needs them.

All 4 AlgoVault tools at a glance

MCPServerAdapter(...) yields 4 CrewAI BaseTool objects when used as a context manager:

Each is invokable directly via tool.run(**kwargs), or assignable to a CrewAI Agent for LLM-driven tool selection.

Optional: wire the tools into a Crew

If you want an LLM to decide when to call which tool, pass the tools list to an Agent and run a Crew:

from crewai import Agent, Crew, Task
from crewai_tools import MCPServerAdapter

server_params = {
    "url": "https://api.algovault.com/mcp",
    "transport": "streamable-http",
}

with MCPServerAdapter(server_params) as tools:
    analyst = Agent(
        role="Crypto market analyst",
        goal="Read AlgoVault's composite verdict for the requested asset and explain it.",
        backstory="A quantitative analyst that calls AlgoVault MCP for cross-venue signal.",
        tools=list(tools),
        verbose=True,
    )
    task = Task(
        description="Get a trade call for BTC on the 4h timeframe and summarise the verdict.",
        expected_output="A one-paragraph summary citing call, confidence, and regime.",
        agent=analyst,
    )
    crew = Crew(agents=[analyst], tasks=[task])
    print(crew.kickoff())

Requires an LLM provider key (OPENAI_API_KEY, ANTHROPIC_API_KEY, etc) plus a model identifier; CrewAI uses litellm under the hood so most providers work. The bare demo above does not.

Why AlgoVault?

• Composite verdict, not raw indicators. One JSON response replaces 26-indicator vote-counting.
• Cross-venue signal. Funding spreads, regime, and sentiment fused across 5 venues – not derivable from any single-venue API.
• Publicly verified. Every signal anchored to Base L2 via Merkle proof. Verify before you subscribe.

90.47% PFE Win Rate · 96,864+ calls · 38+ on-chain batches → view live track record

Next steps

Run the demo at examples/crewai/demo.py · Pair with your CrewAI Crew at https://api.algovault.com/mcp · Verify the track record on https://algovault.com/track-record.

Install (helper plugin)

claude plugin install AlgoVaultLabs/algovault-skills

Once installed, every Skill in the pack is one-line invokable from Claude Code, Cursor, or any MCP-compatible client.

Tutorial © AlgoVault Labs · MIT licensed · Provenance verified 2026-05-18 · crewai 1.14.4 + crewai-tools 1.14.4 (PyPI)