Consensus

Collaborative intelligence, deliberately structured

Better questions deserve more than one answer.

Consensus brings people and AI models into a moderated process that investigates evidence, tests disagreement, and produces a clear synthesis.

Install Consensus Explore the source
  • 18reasoning methods
  • Any mixof humans and AI
  • Local-firstSQLite persistence

The discussion room

Many perspectives. One inspectable process.

Participants contribute in turn while the moderator tracks the thread, surfaces tensions, and builds a running storyboard toward a final conclusion.

Consensus discussion interface showing AI participants, a moderated conversation, and a running storyboard

Choose the structure that fits the question

Reasoning methods, not just chat.

Move from open exploration to disciplined analysis. Each method controls the phases, prompts, validation, and synthesis while the core engine handles the panel.

01

Analysis of Competing Hypotheses

Compare evidence against every plausible explanation and focus on what disconfirms, not what merely supports.

  1. Hypothesise
  2. Gather evidence
  3. Evaluate
  4. Analyse
02

Double Crux

Find the factual belief driving a disagreement and test the point that could change minds.

03

Delphi Method

Collect anonymous estimates, share the distribution, and revise without social pressure.

04

Red Team / Blue Team

Construct, attack, revise, and assess a conclusion through rotating adversarial roles.

05

Weighted Decision Matrix

Rank options against shared criteria and expose whether the winner survives sensitivity analysis.

+13

A full analytical toolkit

Tree of Thoughts, Court of Law, Premortem, Belief State Diffusion, Nominal Group Technique, Recursive Decomposition, and more.

See every method

A serious workspace for inquiry

Models can contribute. They can also investigate.

Give each participant the context, tools, and provider that fit its role. Consensus keeps the discussion, evidence trail, costs, and final result together.

01

Provider freedom

Mix OpenAI, Anthropic, Ollama, DeepSeek, LM Studio, vLLM, and compatible endpoints.

02

Evidence and documents

Search the web, fetch sources, query documents with RAG, and track grounding by turn.

03

Tools that extend

Use built-in Python, vision, and document tools or connect external capabilities through MCP.

04

Memory with boundaries

Optional per-participant memory, semantic retrieval, and configurable context strategies.

Open source · Python 3.11+

Start a better discussion.

Install the current release, launch the desktop app, then add the model providers and participant profiles you want at the table.

Terminal 
uv tool install consensus-app
Then run consensus