Tools Reference

The plugin registers two tools in OpenClaw: axiom_consensus for full multi-model consensus and axiom_verify for fact verification.

axiom_consensus

Run multi-model AI consensus via the Axiom Protocol. Sends a query to multiple LLMs and iteratively converges their responses using φ-convergence.

Parameters

ParameterTypeRequiredDescription
querystringYesThe question or statement to achieve consensus on.
modelsstring[]NoArray of model identifiers (min 2). Defaults to config defaultModels.
structurestringNoConsensus structure: fugue, sonata, or concert. Defaults to config structure.
max_iterationsnumberNoMax convergence iterations (1–7). Defaults to config maxIterations.

Examples

Simple query with defaults:

OpenClaw
axiom_consensus query="What is the capital of France?"

Custom models and structure:

OpenClaw
axiom_consensus query="Should we use SQL or NoSQL for this project?"
  structure="sonata"
  models=["openai/gpt-5.4", "anthropic/claude-sonnet-4-6", "google/gemini-2.5-flash"]

Quick consensus with fewer iterations:

OpenClaw
axiom_consensus query="Is Python dynamically typed?"
  max_iterations=3

Response Format

The tool returns a formatted result containing:

  • D-score — divergence between models (0 = perfect consensus)
  • Confidence — (1 - D) × 100%
  • Iterations used — how many rounds were needed
  • Structure — which consensus pattern was used
  • Model positions — each model's response and distance from consensus
  • Convergence path — D-score at each iteration

The result also includes a details object with structured data:

Response details
{
  "consensus_reached": true,
  "d_score": 0.032,
  "confidence": "96.8%",
  "iterations": 4,
  "structure": "fugue"
}

axiom_verify

Verify a factual claim via multi-model consensus. The tool wraps the claim in a verification prompt and uses the sonata structure (thesis → antithesis → synthesis) with 5 fixed iterations.

Parameters

ParameterTypeRequiredDescription
claimstringYesThe factual statement to verify.
modelsstring[]NoModel identifiers (min 2). Defaults to config defaultModels.

Examples

OpenClaw
axiom_verify claim="Python was created by Guido van Rossum in 1991"
OpenClaw
axiom_verify claim="The speed of light is approximately 300,000 km/s"
  models=["openai/gpt-5.4", "anthropic/claude-sonnet-4-6"]

Response Format

The verification tool returns a verdict with:

  • Verdict — TRUE, FALSE, or UNCERTAIN
  • D-score — agreement level between models
  • Confidence — (1 - D) × 100%
  • Models queried — number of models used
  • Iterations — always 5 for verification

How verification works internally

The plugin wraps your claim in a prompt: "Verify this claim. Respond with TRUE, FALSE, or UNCERTAIN, followed by a concise explanation with evidence." It then runs sonata consensus to get multi-model agreement on the verdict.

D-Score Quick Reference

D-Score RangeLevelConfidence
0.00 – 0.05Strong consensus95–100%
0.05 – 0.20High confidence80–95%
0.20 – 0.40Moderate60–80%
0.40 – 0.60Low confidence40–60%
0.60 – 1.00No consensus0–40%