Python + MCP quickstart

How do you build a crypto trading agent in Python?

Q: How do I call AlgoVault from Python?

Use an MCP client. Connect to the Streamable HTTP endpoint, initialize a session, then call the get_trade_call tool with a coin, timeframe, and venue. The free tier needs no API key.

Q: Does AlgoVault execute trades or only return calls?

It only returns calls. AlgoVault is the call layer your agent queries, not the agent framework or the executor. It does not place orders, hold custody, or size positions. Your code does that.

Q: What do I still build?

Execution against your exchange or broker, position sizing, risk limits, and the agent loop that decides when to call and when to act. AlgoVault supplies the interpreted verdict; you supply the rest.

Q: Is there a free dev tier?

Yes. The free tier includes a monthly call allowance, HOLD calls are always free, and per-call micropayments are supported for higher volume. No API key needed to start.

Build a Python agent by wiring an MCP client to AlgoVault for trade calls. One call returns a composite verdict — direction, confidence, regime — with a 91.6% PFE win rate over 233,000 Merkle-anchored calls. Your agent reads the verdict and acts.

Scope: AlgoVault is the call layer your agent queries, not the agent framework or the executor. You call get_trade_call for the verdict; your code handles execution, sizing, and risk.

How do I call AlgoVault from Python?

Connect an MCP client to the Streamable HTTP endpoint, then call one tool. The free tier needs no API key.

# pip install mcp
from mcp import ClientSession
from mcp.client.streamable_http import streamablehttp_client

async with streamablehttp_client("https://api.algovault.com/mcp") as (r, w, _):
    async with ClientSession(r, w) as s:
        await s.initialize()
        res = await s.call_tool("get_trade_call",
                                {"coin": "BTC", "timeframe": "1h"})
        # res -> {"call": "BUY", "confidence": 72, "regime": "TRENDING_UP", ...}

What does AlgoVault handle, and what do you?

AlgoVault returns the interpreted verdict: direction, confidence, regime, and factors across 5 venues.
You build the agent loop, execution against your exchange or broker, position sizing, and risk limits.

This split keeps your agent thin: no factor engineering, no per-venue normalization, no track-record bookkeeping. See the Model Context Protocol spec and the docs for client setup.

How do you verify the calls your agent acts on?

Every call is hashed and published on Base L2 before its outcome is known. Recompute any call's hash and check it against the on-chain Merkle root at /verify. The live win rate is at /track-record.

When is it the wrong tool?

If you want to engineer your own factors and own the model, AlgoVault is not for you. It is for builders who want a maintained, verifiable verdict to call, not a trading desk to run. It returns calls; it never places orders.

FAQ

How do I call AlgoVault from Python?

An MCP client to the Streamable HTTP endpoint, then get_trade_call with a coin and timeframe. No API key on the free tier.

Does AlgoVault execute trades or only return calls?

Only returns calls. It is the call layer, not the executor; your code handles orders, sizing, and risk.

What do I still build?

Execution, sizing, risk limits, and the agent loop. AlgoVault supplies the verdict.

Is there a free dev tier?

Yes — a monthly call allowance, HOLD calls always free, per-call micropayments for higher volume.

Verify it yourself

See the live track record, verify any call on-chain at /verify, or read how it works. Built by AlgoVault Labs.

Run get_trade_call free — 100 calls/month → Read the Python setup →

This is call interpretation, not investment advice; agents decide execution.