Join the conversation with the Mangrove community on Discord. Join Discord →

The engine behind algorithmic trading strategies.

Generate. Backtest. Sieve. Execute.

Build your own automated trading system, develop thousands of strategies and score them with our predictive sieve model without wasting resources on countless backtests, and run them with deterministic execution and built in guardrails, all behind one API key.

Mangrove API

Strategy Generation

Risk Management

Deterministic Execution

quickstart.py
$ pip install mangroveai
$ export MANGROVE_API_KEY=sk_live_…

from mangroveai import Mangrove
mg = Mangrove()

# describe a thesis in plain language
s = mg.generate("momentum breakout on AVAX")

# ~30ms ML pre-flight before you spend compute
mg.sieve(s)
{ "winning": 0.6977 }  ← worth your time

Trade your Kraken account: two ways to connect.

Kraken Connect

OAuth, ephemeral keys.

Link your account with OAuth; Mangrove mints ephemeral, scoped keys and never holds a static credential.

Bring your own key

Client-side signing.

Your API key stays on your machine, signs client-side, and talks straight to Kraken.

Developer tools

mangrove-agent

Production-ready AI trading agent designed to use the Mangrove API and MCP servers. Clone, set your key, deploy.

mangroveai

Python SDK for strategies, signals, backtesting, and market data.

pip install mangroveai

mangrovemarkets

DEX aggregation, wallet, and portfolio SDK for agents.

pip install mangrovemarkets

Join the community on Discord

Get help, share strategies, and shape the roadmap.

From idea to live execution, one pipeline.

Mangrove is not a data bundle with a strategy layer bolted on. The engine is the product. Every step, from generation to execution, is exposed through the same API. And underneath it all: build, test, and operate with a developer-first workflow: AI copilot guidance, integrated knowledge and sentiment tooling, and a selective data access model built for development teams.

01

Generate

Describe a thesis in plain language and the AI copilot assembles a valid strategy, with entries, exits and parameters, from a deep library of signals, indicators, and candlestick patterns.

Natural language to strategy

Copilot-guided iteration

223 signals · 99 indicators · 27 patterns

02

Backtest

Run each candidate across millions of historical bars. Backtests are metered by the candlesticks they process, so a daily-bar scan and a three-year five-minute deep-dive are priced by what they actually compute.

Millions of historical bars

Metered by candlesticks, not runs

Any timeframe, any asset

03

Sieve

Before you spend compute, the Sieve pre-flights every strategy. Trained on over 1.24M strategies, it predicts which ones are worth running and which are noise. 5 out of 6 fail a backtest.

ML pre-flight in ~30 ms

Trained on 1.24M+ strategies

Skip strategies that will not trade

04

Execute

Promote a proven strategy to deterministic execution with built-in risk controls, so live behavior tracks the backtest instead of drifting away from it.

Deterministic, reproducible execution

Built-in risk controls

Backtest-to-live consistency

Sift the strategies. Skip the noise.

The pre-flight check for every backtest.

5 out of 6 strategies fail a backtest. The Mangrove sieve is trained on a dataset of over 1.24M strategies and lets you know which ones are worth your time, and which ones aren’t.

sieve / score
$ curl -X POST https://api.mangrovedeveloper.ai/api/v1/oracle/sieve/score \
    -H "Authorization: Bearer $TOKEN" \
    -d '{"strategies": [{
          "asset": "AVAX",
          "entry": [
            {"name": "bb_upper_breakout",
             "signal_type": "TRIGGER",
             "params": {"window": 95}}
          ],
          "exit": [{"name": "macd_bearish_cross", ...}]
        }]}'

← 200 OK · 30 ms

{
  "predictions": [{
    "binary": {"p_trades": 0.9999, "p_no_trades": 0.0000},
    "four_class": {
      "losing":    0.2426,
      "no_trades": 0.0000,
      "wash":      0.0596,
      "winning":   0.6977   ← worth your time
    }
  }],
  "model_version": "mangrove-sieve:0b9a2da0d827"
}

Built for teams shipping production-grade algorithmic trading infrastructure.

Trading Infrastructure Teams

Build API-driven strategy workflows, backtesting pipelines, and deployment guardrails.

Quant and Research Teams

Combine sentiment, event, and market signals to test hypotheses and improve system quality.

Advisors and Multi-Account Operators

Standardize strategy controls and reporting across multiple accounts and mandates.

Enterprise Product Builders

Integrate bespoke data and execution flows with governance and support requirements.

Purpose-built infrastructure for autonomous AI agents in algorithmic trading.

1

Autonomous Strategy Research

AI agents can independently research market conditions, analyze historical data, and propose strategy modifications without human intervention.

  • RAG-powered knowledge base
  • Real-time market data access
  • Automated signal discovery

2

Multi-Agent Collaboration

Multiple AI agents can work together, sharing context and coordinating actions across strategy development, backtesting, and optimization.

  • Shared conversation context
  • Parallel backtesting
  • Collaborative optimization

3

Context-Aware Decision Making

Agents maintain full context across sessions, remembering past strategies, user preferences, and performance metrics for intelligent recommendations.

  • Persistent conversation history
  • Strategy versioning
  • Performance tracking

A production-ready trading agent.

Skip the boilerplate. mangrove-agent is a deployable AI trading bot: FastAPI + MCP, ready for GCP Cloud Run. Clone it, set your API key, and you have an agent wired to the full strategy engine.

deploy
$ git clone github.com/MangroveTechnologies/mangrove-agent
$ cd mangrove-agent
$ export MANGROVE_API_KEY=sk_live_…
$ make deploy      # FastAPI + MCP → GCP Cloud Run

✓ agent live · wired to the strategy engine

10+

Data Sources

15+

Blockchains

30+

Models

100%

API-First

The engine runs on premium feeds.

Licensed on your own, their Pro tiers stack up fast.

CoinGecko Analyst$129 / mo
CoinMarketCap Standard$299 / mo
DeFiLlama Pro$300 / mo
Nansen Pro$69 / mo
Whale Alert$30 / mo
Stacked total$827 / mo

With Mangrove Pro

$29.99/ month

Unified access to every one of them, about $797 / month saved, on top of the strategy engine itself.

Developer API pricing.

The developer API is its own subscription, separate from your Mangrove app subscription. Individual plans are for single users; team plans include seats and are billed at the organization level.

🎁 PROTRIAL. New paid subscribers get 14 days of free Pro with code PROTRIAL at checkout.

Pro

$29.99/mo

The developer API subscription, billed separately from your Mangrove app plan.

  • 20 strategies
  • Up to 500 backtests / month
  • 500,000 tokens / month
  • 1,200 API calls / month
  • Full crypto data access
  • DeFiLlama Pro analytics
  • Social sentiment and news indicators

One subscription. One API key. One Mangrove.

Generate, backtest, sieve, and execute strategies, all behind a single key.