Build on QuantLab

1. Sign in, go to Settings then API, and create a key. Copy it immediately, it is shown once.

2. Every request needs it as a bearer token:

curl https://alorny-quantlab.vercel.app/api/v1/me \
  -H "Authorization: Bearer qlk_your_key_here"

Every response is one of two shapes:

{ "ok": true, "data": { ... } }
{ "ok": false, "error": "some_error_code", "hint": "what to do about it" }

Create a project (a trading bot, indicator, script, panel, or library):

curl https://alorny-quantlab.vercel.app/api/v1/projects \
  -H "Authorization: Bearer qlk_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{"name":"London breakout bot","platform":"mql5","requirements":"MT5 EA that trades the London session breakout with a 1 percent risk stop"}'

Send a build request. This runs QuantLab's own generate and compile pipeline, the same one the in app chat uses, and it consumes the account's build quota:

curl https://alorny-quantlab.vercel.app/api/v1/projects/PROJECT_ID/messages \
  -H "Authorization: Bearer qlk_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{"content":"Add a daily loss limit of 3 percent that stops new trades once hit"}'

Poll the project, then fetch the compiled source once the build is ready:

curl https://alorny-quantlab.vercel.app/api/v1/projects/PROJECT_ID \
  -H "Authorization: Bearer qlk_your_key_here"

curl https://alorny-quantlab.vercel.app/api/v1/builds/BUILD_ID/source \
  -H "Authorization: Bearer qlk_your_key_here"

Run a real MT5 Strategy Tester DeepTest on the account's signed-in QuantLab desktop app (mode can also be preview for a quick estimate):

curl https://alorny-quantlab.vercel.app/api/v1/projects/PROJECT_ID/backtests \
  -H "Authorization: Bearer qlk_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{"mode":"real","symbol":"EURUSD","timeframe":"M15","from":"2023-01-01","to":"2024-12-31","deposit":10000}'

Poll for status and metrics:

curl https://alorny-quantlab.vercel.app/api/v1/backtests/BACKTEST_ID \
  -H "Authorization: Bearer qlk_your_key_here"

Once a real backtest is ready, validate the edge with a market permutation test, or pull tester diagnostics if it misbehaved. Neither consumes AI or quota:

curl https://alorny-quantlab.vercel.app/api/v1/backtests/BACKTEST_ID/mcpt \
  -H "Authorization: Bearer qlk_your_key_here" -d '{"n":50}'

curl https://alorny-quantlab.vercel.app/api/v1/backtests/BACKTEST_ID/diagnose \
  -H "Authorization: Bearer qlk_your_key_here"

Download the MCP server, then add it with one command:

curl -O https://alorny-quantlab.vercel.app/dev/quantlab-mcp.mjs
claude mcp add quantlab -e QUANTLAB_API_KEY=qlk_your_key_here -- node quantlab-mcp.mjs

That gives Claude Code these tools: quantlab_list_projects, quantlab_create_project, quantlab_get_project, quantlab_send_build_request, quantlab_get_build_source, quantlab_run_backtest, quantlab_get_backtest, quantlab_run_permutation_test, quantlab_run_diagnostics, quantlab_list_library.

If you keep a project level CLAUDE.md, this paste block gives an agent enough context to use the tools well:

## QuantLab MCP tools

Use the quantlab_* tools to build and test trading bots on QuantLab instead of
writing MQL5/NinjaScript/Pine by hand:
- quantlab_create_project then quantlab_send_build_request to generate and
  compile a strategy using QuantLab's own frameworks
- quantlab_get_project / quantlab_get_build_source to check status and read
  the generated source
- quantlab_run_backtest (mode "real" for a genuine MT5 Strategy Tester
  DeepTest) then quantlab_get_backtest to read metrics
- quantlab_run_permutation_test and quantlab_run_diagnostics to sanity check
  a result before trusting it
All of these consume the connected QuantLab account's normal plan quota.

The MCP server is a single dependency free file (QUANTLAB_API_KEY and optional QUANTLAB_BASE_URL are its only config), so you can also point any other MCP capable agent at it the same way.