OneChater CLI

A multi-model AI terminal. Connect the models you already pay for, and they think together: every model answers, then one synthesizes a single, better reply. With persistent memory and workspace tools, it works like an agent — right in your terminal.

BYOK — your API keys never leave your machine (~/.onechater/). The CLI talks to each provider's API directly. No OneChater account, no proxy.

npm install -g onechater
onechater

Or run it without installing:

npx onechater

(och is a shorter alias for onechater.)

Getting started

Launch it and connect a model — the menus are interactive (↑↓ + enter), so you never type provider names:

onechater

› /providers        ← pick a provider, paste its key, toggle on/off

Free keys (no credit card) are available for Groq, Google and OpenRouter — the menu links them. Connect two or more models and fusion turns on automatically.

Fusion — models that think together

When 2+ models are connected, every message fans out to all of them in parallel. The first model then synthesizes their answers into one coherent response, keeping the best of each and dropping the contradictions. One model connected = a normal single-model chat.

Workspace tools

Ask it to build or change things and it acts like Claude Code / Cursor / Aider — by requesting tools, never touching your system directly. Every action is validated, confined to the current folder (the workspace), and you approve it:

OneChater wants to:
  • Create file: app/page.tsx
  • Run command: npm install
[a]llow · [d]eny

• File access can't escape the workspace; dangerous commands are blocked outright.
• Edits: create_file, write_file (full rewrite), edit_file (precise snippet swap), delete_file, create_folder.
• Read & search: read_file, list_directory, find_files (glob), grep (regex).
• Git: git_status, git_diff, git_commit, git_checkout.
• Other: run_command, load_memory, remember.
• Every action is logged to ~/.onechater/audit.log (view recent ones with /audit).

Memory

Tell it something about you ("remember my name is…") and it saves a persistent profile to ~/.onechater/memory.md. That memory is injected into every model on every session, so any of them already knows you. View it with /memory.

Slash commands

Type / for a navigable command palette (↑↓ to move, enter to pick):

| Command | What it does | |---|---| | /providers | Add a key, toggle models on/off, favorite, see the current mode | | /disconnect | Remove a model | | /model | Set a provider's model | | /default | Pick the synthesizer (and default model) | | /tools | List the workspace tools | | /workspace | Show the workspace directory | | /audit | Recent tool actions log | | /memory | Show the persistent memory | | /clear | Reset the conversation | | /help | Command list | | /exit | Quit |

Shortcuts: Tab fills the highlighted command · double-Esc brings your last message back to edit and resend · Ctrl-C / Ctrl-D exit.

Providers

OpenAI · Anthropic · Google · Groq · OpenRouter · xAI · Mistral · DeepSeek

Free tiers: Groq, Google, OpenRouter.

Privacy

• Keys live in ~/.onechater/config.json (locked to 0600 on POSIX). Nothing is proxied through or stored on a OneChater server.
• Conversations go straight from your machine to the provider you connected.
• Memory and config are plain local files you can read, edit, or delete.

Requirements

Node.js ≥ 18.

© OneChater. All rights reserved.