hermes - 💡(How to fix) Fix [Integration] SIGMA Guard MCP server: deterministic verification before agent output

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
RAW_BUFFERClick to expand / collapse

Hermes Agent supports MCP servers. I built one that adds a deterministic verification layer before agent output.

It is live and installable now:

pip install sigma-guard[mcp]
sigma-guard-mcp

Three tools:

verify_claims   - check LLM claims for structural contradictions
verify_graph    - verify a full knowledge graph
check_write     - test a proposed graph write before commit

Hermes MCP config (examples/hermes_mcp_config.json in the repo):

{
    "mcpServers": {
        "sigma-guard": {
            "command": "sigma-guard-mcp",
            "args": []
        }
    }
}

The use case for Hermes specifically:

Hermes has persistent memory that accumulates across sessions. Over time, contradictions build up silently. One session stores "customer prefers annual billing." A later session stores "customer rejected annual billing." Both persist. Both get retrieved into context. The agent uses conflicting facts without knowing they conflict.

SIGMA Guard catches that before the agent emits its answer.

This is not an LLM judging another LLM. It uses sheaf cohomology (deterministic math) to check whether claims can all be true together. If they cannot, it returns the exact contradiction and a cryptographic proof receipt.

Example output from verify_claims:

{
    "verdict": "UNSAFE",
    "safe_to_rely_on": false,
    "claims_checked": 2,
    "contradiction_count": 1,
    "contradictions": [{
        "severity": "CRITICAL",
        "location": ["Component_X (shipping)", "Component_X (production)"],
        "explanation": "disagree on: component_status, timeline"
    }],
    "receipt_id": "sigma:proof:65e1139b...",
    "elapsed_ms": 1.92,
    "deterministic": true
}

The agent generates. The MCP server verifies. The agent decides whether to emit, revise, or flag.

No model modification required. No fork. No training. Just add an MCP server that returns signed verdicts.

Repo: https://github.com/Jasonleonardvolk/sigma-guard MCP docs: docs/mcp_server.md LLM integration guide: docs/llm_agent_integration.md Hermes config example: examples/hermes_mcp_config.json

Try the verification demo without MCP in 30 seconds:

git clone https://github.com/Jasonleonardvolk/sigma-guard
cd sigma-guard
pip install -e .
python examples/verify_llm_output.py

Happy to help with integration if there is interest.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

hermes - 💡(How to fix) Fix [Integration] SIGMA Guard MCP server: deterministic verification before agent output