Observed since 20 Oct, 17:43 UTC
No superseded metadata states collected. Unchanged observations do not create new revisions.
Multi-agent coding CLI combining Claude, Gemini, and Codex into a single TUI
A TUI coding agent that combines Claude, Gemini, and Codex into a single CLI. Uses your locally installed coding agents as the runtime — no API keys needed.
┌─────────────────────────────────────────┐
│ 0mni v0.1.0 · ● claude-sonnet-4-6 │
├─────────────────────────────────────────┤
│ │
│ You: Fix the login bug in auth.ts │
│ │
│ ● Claude: I'll look at auth.ts... │
│ → read_file src/auth.ts │
│ → edit_file src/auth.ts │
│ │
│ Done. Fixed the null check on line 42. │
│ │
├─────────────────────────────────────────┤
│ > _ │
└─────────────────────────────────────────┘
npm install -g 0mni-cli
Or run directly:
npx 0mni-cli "fix the failing tests"
You need at least one of these coding agents installed and authenticated:
npm install -g @anthropic-ai/claude-codenpm install -g @google/gemini-clinpm install -g @openai/codex0mni uses your existing CLI authentication. No separate API keys required.
# Interactive TUI
0mni
# Start with a prompt
0mni "explain this codebase"
# Force a specific provider
0mni --provider claude "fix the bug"
0mni -p gemini "search for the error"
0mni -p codex "refactor the auth module"
# Auto-route to best provider per task
0mni --auto "find and fix the performance issue"
# Specify a model
0mni -m claude-opus-4-6 "design the new API"
# Auto-approve all tool calls
0mni --yolo "fix all lint errors"
Inside the TUI, use these commands:
| Command | Description |
|---|---|
/claude | Switch to Claude |
/gemini | Switch to Gemini |
/codex | Switch to Codex |
/auto | Toggle auto-routing |
/model <name> | Set a specific model |
/clear | Clear conversation |
/cost | Show token usage and cost |
/help | Show all commands |
When --auto is enabled, 0mni classifies your task and routes to the best provider:
| Task Type | Provider | Why |
|---|---|---|
| Complex reasoning, debugging | Claude | Best at multi-step reasoning |
| Code generation, review | Claude | Strong code understanding |
| Search, web queries | Gemini | Built-in Google Search grounding |
| Multimodal (images) | Gemini | Best multimodal support |
| Refactoring | Codex | Optimized for code transforms |
| Simple Q&A | Claude Haiku | Fast and cheap |
0mni is a TUI wrapper that spawns your installed coding agents as subprocesses:
0mni TUI (React + Ink)
│
├── claude -p "prompt" --output-format stream-json
├── gemini -p "prompt" --output-format stream-json
└── codex exec "prompt" --json
Each CLI handles its own:
0mni parses the streaming JSON output and displays it in a unified terminal UI.
0mni supports skills.sh for extensibility:
# Install skills
0mni skills add vercel-labs/agent-skills
0mni skills add anthropic/skills
# List installed skills
0mni skills list
Create .0mni/config.toml in your project or ~/.config/0mni/config.toml globally:
defaultProvider = "claude"
autoRoute = false
yolo = false
[models]
claude = "claude-sonnet-4-6"
gemini = "gemini-2.5-flash"
codex = "gpt-5.4"
MIT
Download counts and popularity will appear here as the collector receives them.
| Version | Published | Direct dependencies | Open version |
|---|---|---|---|
| 0.1.1Registry latest | 20 Mar 2026 | 18 |
This describes our collected data; it is not a security assessment.
Observed since 20 Oct, 17:43 UTC
No superseded metadata states collected. Unchanged observations do not create new revisions.
package:db6b8a0314d65e2fa1b220540ea4ff24f91d24d37548b309de5313752a16f7bb