Extract includes, classes, structs, enums, functions, methods, event handlers, and calls.
OPEN SOURCE · PUBLIC BETA
MQL5 CodeGraph
Evidence-backed static code-graph intelligence for MQL5 repositories — local, deterministic, and usable from a CLI, dashboard, or AI agent.
WHY IT EXISTS
AI needs sourced context
MQL5 CodeGraph reads .mq5 and .mqh files in an operator-trusted repository, extracts structure and relationships, and produces a queryable JSON snapshot. Evidence origins stay explicit so an agent cannot quietly promote inference into fact.
13read-only MCP tools
5workflow skills
1.0.0intelligence contract
CAPABILITIES
One kernel, several surfaces
Query symbols, context, impact, directed paths, and GraphML exports.
Use a loopback dashboard for search, graph navigation, and source evidence.
Connect an experimental local MCP surface with 13 read-only tools and five Codex workflow skills.
INSTALL V0.3.0
Python 3.11+. No cloud account required.
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"TRUST BOUNDARY
Be explicit about what it cannot prove
- Designed for trusted local repositories and a single-user workflow.
- Not a MetaEditor compiler, security sandbox, or proof of runtime/trading behavior.
- MCP snapshots live in RAM and require re-indexing after source changes or process restart.
- Graphify is optional; remote processing occurs only with explicit operator authorization.
Design in public
Discuss architecture in Discussions or report reproducible defects through Issues.
Contribute responsibly
Read the contribution guide and use a private advisory for vulnerabilities.