Senteron v0.1

A multi-LLM consensus engine. Sends one prompt to up to ten language models, runs peer-review convergence between them, and returns a single synthesized answer — with full provenance back to which model said what.

What it does

Single-model answers vary, hallucinate, and confidently miss obvious points. Senteron dispatches the same prompt to Claude, GPT, Gemini, and others in parallel, then runs a configurable convergence pipeline:

• Dispatch. All models answer in parallel.
• Peer review. Each model scores the others' answers on criteria the prompt itself implies.
• Head-to-head. Top responses face off until one wins on the merits.
• Synthesis. A judge model writes the final answer, drawing the strongest parts from each candidate.

The result is one answer with measurably less variance and fewer single-model blind spots than any of its inputs.

Who it's for

Power users who already feel the limits of single-model output: researchers comparing model behavior, prompt engineers stress-testing formulations, decision-makers who want a second (and third, and fourth) opinion before acting on a generated answer. Requires API keys from at least two providers; not a replacement for a hosted chat product.

How to install

Senteron runs as either a CLI tool or an MCP server (for Claude Desktop and other MCP clients). Both ship from the same repo.

See the documentation for the install recipe, Claude Desktop bundle, and smoke-test sequence.

Status

Early public preview. The CLI is stable; the MCP server has shipped a Claude Desktop bundle. Issue tracker, roadmap, and source live on GitHub.