lm-oandaproxy

Oanda API proxy with unlimited historical candle depth.

Continuously captures OHLCV candle data from Oanda, stores it as compressed archives, and serves it via an Oanda v3-compatible REST API. Any Oanda client can point at this proxy and get years of historical data in a single request — no 5000-bar limit.

All other Oanda endpoints (accounts, orders, trades, positions, pricing, streaming) are transparently proxied to the real Oanda API.

Hosted Service

A public instance is available at https://oanda.langmart.ai — use it directly without any deployment.

Just swap your Oanda base URL and use your own Oanda Bearer token:

# Health check (no auth required)
curl https://oanda.langmart.ai/health

# Candles — use your own Oanda token
curl -H "Authorization: Bearer YOUR_OANDA_TOKEN" \
  "https://oanda.langmart.ai/v3/instruments/EUR_USD/candles?granularity=H4&count=5000"

# All Oanda endpoints work — accounts, orders, trades, pricing, streaming
curl -H "Authorization: Bearer YOUR_OANDA_TOKEN" \
  "https://oanda.langmart.ai/v3/accounts"

Both practice and live Oanda tokens work. Your token determines everything:

| Your token type | Candle history | Live fill | Orders, trades, accounts | |----------------|---------------|-----------|-------------------------| | Practice | Shared archive | Fetched from api-fxpractice.oanda.com | Proxied to api-fxpractice.oanda.com | | Live | Shared archive | Fetched from api-fxtrade.oanda.com | Proxied to api-fxtrade.oanda.com |

Candle history is shared because Oanda practice and live return identical OHLCV market data — only account-specific endpoints (orders, trades, positions, balance) differ between environments.

The hosted instance captures data for major forex pairs, commodities, and metals across 8 timeframes (S5 through H8).

Self-Hosted Setup

Install

npm install -g lm-oandaproxy

Or run directly:

npx lm-oandaproxy

Requires Node.js 18+.

Quick Start

# 1. Login with your Oanda API token
lm-oandaproxy login

# 2. Start the capture daemon (fetches candle data in background)
lm-oandaproxy capture start

# 3. Start the API server
lm-oandaproxy server start

# 4. Use it — same as the real Oanda API
curl -H "Authorization: Bearer YOUR_TOKEN" \
  http://localhost:8900/v3/instruments/EUR_USD/candles?granularity=H4&count=5000

Point any Oanda client at http://localhost:8900 instead of https://api-fxpractice.oanda.com.

Install as System Service

# Install systemd services (auto-restart on crash, auto-start on reboot)
lm-oandaproxy install-service

# Remove services
lm-oandaproxy remove-service

This creates and enables two systemd services:

• lm-oandaproxy-server — API server (port 8900)
• lm-oandaproxy-capture — background data capture daemon

Both auto-detect the current Node.js path, working directory, user, and port configuration. Requires Linux with systemd and sudo access.

After installing, manage with standard systemctl:

sudo systemctl status lm-oandaproxy-server
sudo systemctl restart lm-oandaproxy-capture
journalctl -u lm-oandaproxy-server -f

Updating

cd /path/to/lm-oandaproxy
git pull && npm run build
sudo systemctl restart lm-oandaproxy-server lm-oandaproxy-capture

How It Works

                          lm-oandaproxy (port 8900)
                         ┌─────────────────────────┐
  Client request         │                         │
  GET /v3/.../candles ──>│  Stored data (gzip)     │
                         │  + live Oanda fill      │──> Response (unlimited bars)
                         │                         │
  GET /v3/accounts/... ─>│  Proxy to real Oanda    │──> Response (passthrough)
  POST /v3/.../orders ──>│  (practice or live)     │──> Response (passthrough)
                         └─────────────────────────┘
                                    ^
                         ┌──────────┴──────────┐
                         │  Capture Daemon      │
                         │  S5 every 3h         │
                         │  M1 every 12h        │
                         │  M5/M15 every 3d     │
                         │  M30-H8 every 14d    │
                         └─────────────────────┘

Candle requests are served from the compressed local archive, merged with a live Oanda fetch to fill the gap to the current bar. The client always gets the most recent data.

Everything else (accounts, orders, trades, positions, pricing, streaming) is proxied directly to the real Oanda API. POST, PUT, PATCH, DELETE all work.

CLI Reference

lm-oandaproxy <command>

  login [--practice|--live]              Verify and save Oanda API credentials
  logout [practice|live|all]             Remove saved credentials
  status                                 Show credentials, services, storage

  server start [--port N] [--public]     Start API server (background)
  server stop                            Stop API server
  server foreground [--port N] [--public] Run server in foreground

  capture start                          Start capture daemon (background)
  capture stop                           Stop capture daemon
  capture run-once                       Single fetch cycle (foreground)

  install-service                        Install systemd services (Linux, requires sudo)
  remove-service                         Stop and remove systemd services

  version                                Show version

API Endpoints

Local (stored data + live fill)

| Method | Endpoint | Description | |--------|----------|-------------| | GET | /v3/instruments/{instrument}/candles | Candle data (no 5000-bar limit for from+to queries) | | GET | /v3/accounts/{id}/instruments/{instrument}/candles | Same, account-scoped | | GET | /v3/instruments | List instruments with available granularities | | GET | /v3/accounts/{id}/instruments | Same, account-scoped |

Proxied to Oanda

| Method | Endpoint | Description | |--------|----------|-------------| | GET | /v3/accounts | List accounts | | GET | /v3/accounts/{id}/summary | Account summary | | POST | /v3/accounts/{id}/orders | Create order | | PUT | /v3/accounts/{id}/orders/{id}/cancel | Cancel order | | GET | /v3/accounts/{id}/openTrades | List open trades | | PUT | /v3/accounts/{id}/trades/{id}/close | Close trade | | GET | /v3/accounts/{id}/pricing | Live pricing | | GET | /v3/accounts/{id}/pricing/stream | Price stream (chunked) | | ... | /v3/* | All other Oanda v3 endpoints |

All 32 Oanda v3 endpoints are covered.

Server-only

| Method | Endpoint | Auth | Description | |--------|----------|------|-------------| | GET | /health | Public | Health check | | GET | /status | Public | Storage stats, per-instrument detail, disk usage |

Authentication

Use the same Authorization header as the real Oanda API:

Authorization: Bearer YOUR_OANDA_TOKEN

The server auto-detects your token type and routes accordingly:

• Practice tokens (two hex halves joined by hyphen) route to api-fxpractice.oanda.com
• Live tokens (different format) route to api-fxtrade.oanda.com

This means you can use a single proxy endpoint for both practice and live trading. Candle history is shared (same market data), while account operations (orders, trades, positions) go to the correct Oanda environment based on your token.

No registration or API key needed for the proxy itself — just your existing Oanda token.

Use --public flag to disable auth entirely (for local development only).

Credentials

Dev mode — .env file in project root:

OANDA_API_TOKEN=your-token-here
OANDA_ACCOUNT_ID=101-003-XXXXXXX-XXX

Prod mode — saved via CLI:

lm-oandaproxy login
# Verifies token with Oanda, saves to ~/.lm-oandaproxy/config.json

Data Storage

Candle data is stored as gzip-compressed JSONL (raw Oanda format, no transformation):

~/.lm-oandaproxy/
  config.json              # Saved credentials
  data/
    {INSTRUMENT}/
      S5.jsonl.gz          # 5-second candles
      M1.jsonl.gz          # 1-minute candles
      ...
      H8.jsonl.gz          # 8-hour candles
    .state.json            # Last fetch state per instrument/TF
    logs/
      capture-YYYY-MM-DD.log
      server-YYYY-MM-DD.log
      capture-YYYY-MM-DD.log.gz   # Old logs compressed, never deleted

Timeframes Captured

| TF | Poll Interval | 5000 bars covers | |----|--------------|-----------------| | S5 | 3 hours | 6.9 hours | | M1 | 12 hours | 3.5 days | | M5 | 3 days | 17 days | | M15 | 3 days | 52 days | | M30 | 14 days | 104 days | | H1 | 14 days | 208 days | | H4 | 14 days | 833 days | | H8 | 14 days | 1667 days |

Compression

Raw Oanda JSON is stored at ~135 bytes/bar, compressed to ~12-21 bytes/bar (7-11x ratio). Monthly growth for 10 instruments across all 8 timeframes: ~21 MB compressed.

S5 data has a rolling window cap of 2M bars (~9 months) to bound memory usage during merges.

Programmatic API

import { createServer, runDaemon, loadConfig } from 'lm-oandaproxy';

// Start server programmatically
const server = createServer(8900);
server.listen(8900);

// Or run capture daemon
await runDaemon();

Log Management

• Daily log files: capture-{date}.log, server-{date}.log
• Active log capped at 10 MB (oldest half truncated)
• Logs older than 2 days are compressed to .log.gz
• Compressed logs are kept forever (never deleted)

License

AGPL-3.0-or-later