PM AI CLI

Universal AI terminal CLI — Ollama, Groq, OpenRouter, Google Gemini, Kimi, MiniMax, DeepSeek.

The CLI workflow is the same for all providers: set config → run pm.
If the logo doesn’t render, the CLI still works normally.

1) Install

npm install -g pm-ai-cli

Verify installation:

pm --version
pm --help

2) First run (recommended step-by-step)

Step A — Start setup wizard

pm config setup

This wizard guides you to:

• pick your provider
• paste the appropriate API key
• (optionally) set a default model

Step B — Confirm config exists

pm config show

Step C — Run your first message

pm "Hello world"

Step D — Start interactive chat (menu + slash commands)

pm

3) Provider selection (how to switch quickly)

CLI options you can reuse:

• -p, --provider <provider>: groq | openrouter | ollama | google | kimi | minimax | deepseek
• -m, --model <model>: provider model name
• -s, --system <prompt>: system prompt override

Examples

# Groq
pm -p groq -m <model> "Hello"

# Ollama (local)
pm -p ollama -m <model> "Hello"

# OpenRouter (shortcut)
pm or deepseek/deepseek-r1:free

# OpenRouter with system prompt
pm -p openrouter -m "deepseek/deepseek-r1:free" -s "Be concise" "Explain Docker"

4) Full command list (with examples)

Help / version

pm --help
pm -v

One-shot message (default provider/model)

pm "Explain quantum computing"

Interactive chat

pm
pm chat

Provider shortcut commands

pm groq "hello"
pm deepseek "explain Docker"
pm kimi "write a poem"

# Ollama (local)
pm ollama
# or: pm ollama "your message"

Models (list available models)

pm models
pm models <filter>

Pull (Ollama model)

pm pull llama3.2

Status (provider status)

pm status

Config commands

pm config show
pm config set <key> <value>
pm config setup

Common keys (depending on provider):

• provider
• model
• <provider>-key (e.g., openrouter-key, groq-key, google-key, etc.)

OpenRouter shortcut

pm or <openrouter-model>
pm or deepseek/deepseek-r1:free
pm -p openrouter -m "deepseek/deepseek-r1:free" "Hello"

5) Chat menu + slash commands

Open the interactive menu

In interactive chat, type / and press Enter.

Slash commands (quick reference)

/orm                  Open model switcher
/upload <path>       Upload a file (image/pdf/docx/xlsx/zip/code...)
/retry                Retry last message
/save [file]         Save conversation to a file
/tokens               Show token usage estimate
/system <msg>        Update system prompt
/help                 Show all commands
/exit                 Quit

6) Building from source (contributors)

npm ci
npm run build
node dist/index.js --help

7) Publish (maintainers)

The project is published automatically by GitHub Actions on tags v*.

1. Bump version
2. Create tag vX.Y.Z
3. Push tag (publish workflow runs)