SQXtreme — StrategyQuant X MCP control layer
Wrapper/extension over sq-mcp that adds Windows-first DX, Spanish localization, multi-client configs, and PyPI publish infra for AI agents (opencode, Claude Code, Cursor, VS Code).
Repo: https://github.com/herjarsa/sqxtreme-mcp
What ships in v0.5.0
sq-mcp (DAVIDAROCA27) is the foundation: 482 runtime tools, MIT, 1571 tests. SQXtreme v0.5.0 adds (122 tests: 100 unit + 22 E2E live server, CI verde, OIDC publish):
- Windows installer —
iwr install.ps1 | iex(4 modos: CheckOnly/PyPI/FromSource/FromGit) - Spanish localization — 5 workflow prompts + 4 sub-agentes (
prompts/es/) — es/ - Multi-client configs — Claude Code, opencode, Cursor, VS Code (
examples/) - HTTP transport —
SQXTREME_TRANSPORT=http→streamable-httpen:8765(vs nativo:8082) - 12 features 3 fases — AlgoWizard auto-import, Builder desde frase, Live Watch, Streaming, Data auto-fix, Auto-curator, Ship MT5, Snapshot rollback, Ashare 119, Scheduled Runner, sqx-tools wrapper, Health Dashboard
- CI matrix — Ubuntu + Windows × Python 3.10/3.11/3.12 — ruff + pytest
- PyPI OIDC — Trusted Publishing sin token manual (
publish.yml) - E2E test suite —
tests/e2e/corre contra server vivo en :8765 (22 tests) - 26 audit fixes — seguridad hardened (2026-08-30)
Roadmap (post v0.5.0)
Live trading bridge refinements · Multi-workspace orchestrator · Research Agent Loop (moonshot E1).
Install
# One-liner (after PyPI publish)
iwr https://raw.githubusercontent.com/herjarsa/sqxtreme-mcp/main/install.ps1 -UseBasicParsing | iex
# Or from source
git clone https://github.com/herjarsa/sqxtreme-mcp.git
cd sqxtreme-mcp
.\scripts\install.ps1 -FromSource
Native tools (always available, do not require sq-mcp[full])
These 3 tools are registered directly in server.py (always available, graceful-degradation if sq-mcp missing):
sq_algowizard_import
Import a Java AlgoWizard snippet into the SQ X user/extend/<Type>/ directory.
{
"src_path": "C:\\path\\to\\MyInd.java",
"target_type": "indicators" // optional; inferred from package SQ.<Type> if omitted
}
Returns: {ok, dest_path, snippet_name, snippet_type, warnings} or {ok:false, error, error_type} on validation failure.
Required in .java file: package SQ.<Indicators|Signals|Conditions|...>;.
sq_builder_from_phrase
Parse a Spanish trading phrase and prepare a Builder plan (dry_run by default).
{
"phrase": "Builder para GBPUSD H1 2023.01.01 mean reversion max 5 condiciones",
"dry_run": true // default; if false, would patch project.cfx via cfx_clone + cfx_apply_patch
}
Returns: {ok, dry_run, parsed: {symbol, timeframe, date_from, date_to, max_conditions, strategy_type, raw}, warnings, next_steps}.
sq_live_watch
Compute live position summary metrics from a list of position dicts.
{
"positions": [
{"price": 50.0, "volume": 100, "side": "long"},
{"price": 100.0, "volume": 50, "side": "short"}
]
}
Returns: {net_exposure, gross, pnl_total, n_long, n_short, warnings}. Validates: price must be finite numeric; volume > 0; side ∈ {"long","short"}.
Delegated tools (from sq-mcp, 482+ tools)
When sq-mcp[full] extra is installed, the server delegates all sq-mcp tool groups (pkgutil.iter_modules + import_module + register(mcp)). Key examples:
| Tool | Purpose |
|---|---|
workspace_overview |
Count projects, databanks, .sqx files; mtime of last edits |
workspace_doctor |
Health check: engine + license + databanks + recent errors |
cfx_inspect |
Read .cfx: task count, ranking types, MM methods |
cfx_archetype |
Classify .cfx by task types (Builder, Retester, Optimizer, WalkForward, MonteCarlo) |
cfx_apply_patch |
Patch symbol/timeframe/dateFrom/dateTo across all task XMLs |
cfx_lint |
Validate .cfx structure |
builder_start / builder_stop / builder_watch_progress |
Launch and monitor a Builder run |
retest_strategy |
Retest a single strategy against OOS window |
workspace_insights |
Top 10 strategies by profit_to_dd, symbol/timeframe distribution |
data_health_freshness_gate |
Pre-trade check: data must be <max_age_minutes old (requires last_bar_epoch + now_epoch) |
health_check |
Engine status, license days remaining, fatal errors |
For full list: run server and call MCP tools/list (485+ entries).
Spanish prompts (9)
| Prompt | Purpose |
|---|---|
sqxtreme-morning |
Diagnóstico matutino: engine health + workspace + top strategies + alerts |
sqxtreme-build-and-retest |
Builder desde frase → clonar template → parchear → lintear → start |
sqxtreme-promote |
Evaluar strategy contra gates de riesgo (conservative/balanced/aggressive) |
sqxtreme-post-mortem |
Diagnosticar job fallido + sugerir acciones |
sqxtreme-portfolio-rebalance |
Recomputar pesos + proponer rotación |
agente-constructor |
Sub-agent enfocado en fase de construcción |
agente-promotor |
Sub-agent enfocado en gates de promoción |
agente-diagnostico |
Sub-agent enfocado en diagnosticar fallos |
agente-cartera |
Sub-agent enfocado en portfolio y rebalanceo |
What this gets you (honest scope)
Running an AI agent with SQXtreme + [full] extra: 9 Spanish MCP prompts + 482 sq-mcp runtime tools delegated + 3 native tools + cross-client install + 26 audit security fixes + 22 E2E tests against live server.
NOT included in v0.5.0: a live AI agent actually connected (manual client setup per docs/CLIENTS.md), E2E test with real SQ X engine (file-based tools work, engine-dependent tools require live engine).
Security
- E2E-tested edge cases: phrase injection (
'; DROP TABLE), NaN/Inf prices, empty inputs, traversal attempts - sqxtreme-mcp delegates to upstream sq-mcp which has XXE defense + atomic writes + NaN validation
- 26 audit fixes from v0.4.3 release (security hardened)
- See
CHANGELOG.mdfor full security fix list
License
MIT.
Release files for sqxtreme-mcp 0.5.2
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| sqxtreme_mcp-0.5.2.tar.gz | 55.6 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| sqxtreme_mcp-0.5.2-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 83.4 kB
Release files / sqxtreme_mcp-0.5.2.tar.gz
| Download URL | sqxtreme_mcp-0.5.2.tar.gz |
|---|---|
| Size | 55.6 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
f6e08cad7ea025e786a6436ad517a7059418401228445a4fd473910fc371004c
|
|
BLAKE2b-256 checksum How to use checksums |
73489e1aa2dd0690d24fda15a39ec415305a06f297ccef1ea7f61b309a028268
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Aug 31, 2026.
Transparency logRelease files / sqxtreme_mcp-0.5.2-py3-none-any.whl
| Download URL | sqxtreme_mcp-0.5.2-py3-none-any.whl |
|---|---|
| Size | 27.8 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
4c7a867d4580db9680dfeb2e3df301b39e58691a61b56baa9daf58fc49038a66
|
|
BLAKE2b-256 checksum How to use checksums |
319cba2264ec46d7af98d33a7fb6bfc7cd32ed4570e2949b6e528f6dc2e68e20
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Aug 31, 2026.
Transparency log