Model pools.
One consensus.
Create teams of local LLMs specialized for your stack (TypeScript, Rust, Python…).
They debate, vote, and deliver production‑ready code — 100% private, zero cloud, zero API keys.
Built for developers.
Model Pools
Group models by specialty (TypeScript, Python, Rust…). Auto‑generated, editable system prompts.
Model Selection
Choose which models to use globally or per pool. Visualize your selection anytime.
Code‑First Debates
REASONING + CODE format. Models propose, critique, and refine solutions.
Regenerate Prompts
One command: `aicp pool regenerate-prompt` to refresh any pool's system prompt.
Democratic Voting
Each model votes for the best code — self‑votes are automatically rejected.
Reputation System
SQLite stores accuracy & energy per model. Improves with every debate.
Real‑Time Streaming
Watch every token, argument, and vote as they happen.
100% Local
Runs on Ollama. No data ever leaves your machine.
CLI with nested menus
Six steps. One verdict.
Pull & Select Models
`ollama pull mistral` + `aicp models select`
Create Pool
Name, specialty, and which models participate.
Chat / Debate
Models propose code in REASONING+CODE format.
Argumentation
Each model can critique others' proposals.
Voting
Secret ballot — cannot vote for yourself.
Consensus
Winner writes final answer with syntax highlighting.
From zero to consensus in minutes.
Install, create a pool, and start coding. No cloud, no API keys — just you, your models, and Ollama.
- Select which models to use globally or per pool
- Regenerate any pool's system prompt with `aicp pool regenerate-prompt`
- SQLite reputation system (accuracy & energy)
- Interactive CLI with mouse support
- Syntax‑highlighted final answers
Four workspaces. One framework.
Stop trusting a single model.
Build an expert team.
Open source, MIT licensed. Built for developers who want reliable, private AI assistance.