Directory / dearvn/SOCAgents

dearvn/SOCAgents

Open-source multi-agent AI trading desk for options and futures: dealer gamma, options flow, bull vs bear debate, evidence-cited reports.

Reviewed New (under 90 days old) AI agent 86/100

Not financial advice. A grade measures public repo hygiene only (license, upkeep, tests, paper mode, risk limits, docs). It does not measure returns, safety or suitability. Options trading can lose more than you invest. Read the full policy.

Reviewed. Grok Bot read the code and docs and checked or adjusted each judged factor. Nothing was executed. This is not a human audit. Reviewed on 2026-10-24 against commit d036501, rubric v1.1.

Reviewer summary: Multi-agent options desk that currently produces trade ideas and does not place orders; broker execution is on the roadmap. It has a risk engine (max contracts and max loss), human confirmation hooks and read-only MCP tools. Paper is not applicable until it can place orders. The dry-run code the scanner found belongs to the X reply bot.

Facts

Repository	https://github.com/dearvn/SOCAgents
Type	AI agent
Strategies	Not tagged
Brokers	No supported broker detected
License	Apache-2.0
Language	Python
Last commit	2026-10-22 (3 days before the scan)
Last scanned	2026-10-24
Stars	1 (shown for context, never scored)

Score breakdown

Factor	Points	How it was decided
Risk limits and kill switch	20 / 20	socagents/risk/engine.py enforces max contracts and max loss; external tool calls need confirmation and order verbs are blocked. Reviewer confirmed the scanner result. Scanner evidence (5, automated, can include false matches) `socagents/desk/graph.py:594` [max loss] max_loss_usd=decision.max_loss_usd, `socagents/risk/engine.py:55` [max contracts] max_contracts: int = Field(default=5, gt=0) `socagents/risk/engine.py:54` [max notional or exposure] max_open_risk_usd: float = Field(default=2_000.0, gt=0) `socagents/external_mcp.py:96` [human approval] def needs_confirmation(self) -> bool: `socagents/mcp/server.py:32` [read-only mode] READ_ONLY = ToolAnnotations(
License	15 / 15	Apache-2.0 (OSI approved) Reviewer confirmed the scanner result. Scanner evidence (1, automated, can include false matches) `LICENSE:1` License file
Last commit on default branch	15 / 15	2 days since last commit Fact from the commit date, no judgment involved. Scanner evidence (1, automated, can include false matches) `default branch HEAD` 2026-10-22 (2 days ago)
Tests	15 / 15	369 test cases found in 27 test files; CI test step found Reviewer confirmed the scanner result. Scanner evidence (4, automated, can include false matches) `tests/test_analytics.py:1` 16 test cases `tests/test_cli.py:1` 19 test cases `tests/test_community_provider.py:1` 17 test cases `.github/workflows/ci.yml:34` run: pytest -q
Paper or dry-run mode	N/A	Not applicable. Does not place orders yet (broker execution is on the roadmap) Scanner evidence (2, automated, can include false matches) `docs/README.md:12` \| [X reply bot](x-bot.md) \| answering mentions on X: credentials, cost, caps, dry run, going live \| `socagents/cli/main.py:1338` [dry-run flag] styles = {"replied": "green", "dry_run": "cyan", "skipped": "dim", "failed": "red"}
README quality	8 / 10	install steps, explains what it does, risk disclaimer, 8724 chars Reviewer confirmed the scanner result. Scanner evidence (2, automated, can include false matches) `README.md:1` README with 8724 characters `README.md:69` \| Trade ideas \| educational only (delayed data) \| risk-checked on live data \| v0.1 \|
Broker detection	0 / 10	No broker integration shipped yet. Reviewer confirmed the scanner result.
Total	86 / 100	Not applicable factors are removed and the total is rescaled to 100. Grades: A 85+, B 70+, C 55+, D 40+, F below.

Grades measure public repo hygiene on the scan date. They say nothing about profits, safety or security. We did not run this code. Full rubric and patterns.

Something wrong?

Report a correction Request removal (maintainers)

Corrections and removals are public GitHub issues. We reply within 7 days and every change is in the repo history. Sponsors get no special treatment: see our policy.

Badge for your README

Once the site has a public address, shields.io can read /badge/dearvn--socagents.json from it. Badges show the grade and whether it was reviewed. Maintainers may use one without asking.