GEM²-AI

AI that knows what
it doesn't know.

AI with a built-in truth filter.

Know what to trust→Fix what's wrong→Scale with confidence
Install TPMN CheckerLearn more

Product A — Available now

Try it now

TPMN Checker is an MCP server that evaluates epistemic quality of any LLM output. Install it in Claude Code or Claude Desktop in one command.

Claude Code

npx @anthropic/claude-code mcp add gem2-tpmn -- npx -y gem2-tpmn-mcp

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "gem2-tpmn": {
      "command": "npx",
      "args": ["-y", "gem2-tpmn-mcp"]
    }
  }
}

Docker

docker pull gem2ai/tpmn-checker && docker run --rm -i gem2ai/tpmn-checker

What you get

⊢

Truth Score

0.0–1.0 epistemic reliability score for any content.

⊬

Extrapolation Detection

Flags overclaims and unsupported inferences (EEF).

↻

Auto-Refinement

Strips overclaims and grounds content using source data.

Platform — Coming soon

The full picture

TPMN Checker is just the starting point. The GEM²-AI platform orchestrates multi-agent workflows where every output is truth-filtered, grounded, and auditable.

How it works

1

Orchestrate

GACC routes tasks to specialized AI actors (G, S, D, I, E) with role-specific prompts.

2

Verify

Every output passes through the TPMN truth filter — overclaims are caught before delivery.

3

Ground

Extrapolations are flagged, then grounded with real source data via interpolation.

🔍

Know what to trust

Every claim tagged with epistemic status: grounded, inferred, extrapolated, or unknown.

🔧

Fix what's wrong

Auto-refinement strips overclaims. Interpolation fills gaps with verified data.

🚀

Scale with confidence

Run hundreds of AI agents. Platform-level governance ensures nothing ships unchecked.

Join early access

Be among the first teams to run truth-filtered AI workflows at scale.

Request access

Open standard

Open protocol, governed intelligence

TPMN is an open epistemic notation — anyone can implement it. The GEM²-AI platform is how enterprises run it at scale.

Open

TPMN Specification

  • ✓Epistemic tagging notation (⊢ ⊨ ⊬ ⊥?)
  • ✓Truth filter pipeline (T-P-M-N)
  • ✓Extrapolation detection (EEF)
  • ✓MCP-compatible tooling
Read the spec→
Enterprise

GEM²-AI Platform

  • ✓Multi-agent orchestration (GACC + WMS)
  • ✓Knowledge graph persistence (GEM-Kg)
  • ✓Workflow audit trails & provenance
  • ✓Role-based AI governance
Contact for enterprise→