Install MQL5 CodeGraph v0.3.0 and analyze your first project

Version 0.3.0 is a public beta for Python 3.11 and newer. The core analyzer needs no cloud account or model API key.

1. Install the public wheel

Open PowerShell and run:

python -m pip install "mql5-codegraph[reference,mcp] @ https://github.com/junet03/mql5-codegraph/releases/download/v0.3.0/mql5_codegraph-0.3.0-py3-none-any.whl"

Verify the entry points:

mql5-codegraph --help
mql5-codegraph-mcp

The second command starts a stdio server for an MCP client. It is not an interactive application. You can stop it with Ctrl+C after confirming that the process starts.

2. Select the right project root

Choose a repository containing an EA, indicator, or MQL5 library that you trust. Do not use the mql5-codegraph tool repository itself as the target project. A public URL is not permission to process untrusted source.

Create a snapshot:

mql5-codegraph analyze C:\work\Example-MQL5 --output build/example.codegraph.json --max-work 2000000 --json

--max-work is a bounded analysis budget. If it is exhausted, the tool reports the phase and counters instead of publishing an incomplete graph as if it were complete.

3. Inspect before asking an AI agent

Start with status and a symbol you recognize:

mql5-codegraph status build/example.codegraph.json --json
mql5-codegraph intelligence query build/example.codegraph.json OnTick --json
mql5-codegraph intelligence context-package build/example.codegraph.json OnTick --context-units 40 --json

Read diagnostics and evidence origin. Stored locations describe the analyzed snapshot; if the source changes, create a fresh snapshot.

4. Evidence the graph cannot provide

A static graph does not prove that code compiles. Run the project's MetaEditor workflow and retain compile logs, tests, and backtests separately. CodeGraph helps orientation and evidence linking; compiler and runtime evidence still govern release decisions.