AI编程编辑器

🚀 Installation

⚡ Quick Start

Note: CPE requires a configuration file to define which models and tools to use. There's no zero-config mode—you'll need to set up at least one model before getting started.

Show OpenAI subscription usage

cpe account usage openai

Watch usage live

cpe account usage openai --watch

Override the ChatGPT usage backend base URL

cpe account usage openai --base-url https://chatgpt.com/backend-api

Full Configuration Example

```yaml

📚 Examples

Example Skill Structure

skills/
└── github-issue/
    └── SKILL.md

1. Create a configuration file

Create a cpe.yaml in your project directory or in your user config directory: - macOS: ~/Library/Application Support/cpe/cpe.yaml - Linux: ~/.config/cpe/cpe.yaml - Windows: %AppData%\cpe\cpe.yaml

version: "1.0"

models:
  - ref: sonnet
    display_name: "Claude Sonnet"
    id: claude-sonnet-4-5-20250929
    type: anthropic
    api_key_env: ANTHROPIC_API_KEY  # You choose the env var name
    context_window: 200000
    max_output: 64000
    timeout: 5m
    generationParams:
      temperature: 0.2

Tip: You can quickly add models from the models.dev registry:

> # Add a model from the registry
> cpe config add anthropic/claude-sonnet-4-20250514 --ref sonnet
> 

Use whatever env var name you specified in api_key_env

export ANTHROPIC_API_KEY="your-api-key" ```

⚙️ Configuration

Configuration File Locations

CPE searches for configuration in this order: 1. --config flag (explicit path) 2. ./cpe.yaml (current directory) 3. User config directory: - macOS: ~/Library/Application Support/cpe/cpe.yaml - Linux: ~/.config/cpe/cpe.yaml - Windows: %AppData%\cpe\cpe.yaml

yaml-language-server: $schema=https://raw.githubusercontent.com/spachava753/cpe/refs/heads/main/schema/cpe-config-schema.json

version: "1.0"

Environment Variables

Note: API keys are configured per-model via the api_key_env field. You choose the environment variable name—there are no hardcoded defaults. For example, you could use MY_ANTHROPIC_KEY, OPENAI_API_KEY, or any name you prefer.

List configured models (aliases: models, ls)

cpe model list

Configuring Skills

Skills locations are user-defined in your system prompt template using the {{ skills }} function. There are no hardcoded defaults—you specify exactly which directories to scan:

{{- $skills := skills "./skills" "~/my-custom-skills" "/shared/team-skills" -}}
{{- if $skills }}
<skills>
{{- range $skill := $skills }}
  <skill name={{ printf "%q" $skill.Name }}>
    <description>{{ $skill.Description }}</description>
    <path>{{ $skill.Path }}</path>
  </skill>
{{- end }}
</skills>
{{- end }}

The skills function: - Accepts any number of directory paths - Scans each for subdirectories containing SKILL.md - Returns a list of skill objects (name, description, path) so your template controls the output format (XML, JSON, CSV, etc.)

2. Set your API key

```bash

🔧 CLI Reference

cpe [flags] [prompt]

Core Flags:
  -m, --model string           Specify the model profile to use (required unless CPE_MODEL is set)
  -i, --input strings          Input files or URLs to process
  -n, --new                    Start a new conversation
  -c, --continue string        Continue from a specific conversation ID
  -G, --incognito              Don't save conversation to storage
      --config string          Path to YAML configuration file
      --db-path string         Path to conversation SQLite database
      --skip-stdin             Skip reading from stdin
  -v, --version                Print version and exit

Generation Parameters:
  -t, --temperature float      Sampling temperature (0.0 - 1.0)
  -x, --max-tokens int         Maximum tokens to generate
  -b, --thinking-budget string Budget for reasoning capabilities
      --top-p float            Nucleus sampling parameter (0.0 - 1.0)
      --top-k uint             Top-k sampling parameter
      --frequency-penalty float Frequency penalty (-2.0 - 2.0)
      --presence-penalty float  Presence penalty (-2.0 - 2.0)
      --number-of-responses uint Number of responses to generate
      --timeout string         Request timeout (e.g., '5m', '30s')

Advanced:
      --custom-url string      Custom base URL for the model provider API

Commands:
  account       Manage AI provider accounts (login, logout, usage)
  config        Manage configuration (add, remove models)
  conversation  Manage conversation history [aliases: convo, conv]
                ├─ list    List conversations [alias: ls]
                ├─ print   Print a conversation [aliases: show, view]
                └─ delete  Delete conversations [aliases: rm, remove]
  model         List and inspect models [alias: models]
                ├─ list          List models [alias: ls]
                ├─ info          Show model details
                └─ system-prompt Show rendered system prompt
  mcp           MCP tools
                ├─ list-servers List servers [alias: ls-servers]
                ├─ list-tools   List tools [alias: ls-tools]
                ├─ call-tool    Call a tool
                ├─ info         Server info
                └─ code-desc    Show code mode description
  completion    Generate shell autocompletion scripts (bash, zsh, fish, powershell)

MCP Tool Integration

Connect external tools via the Model Context Protocol. CPE supports three MCP server types:

The bundled text_edit file editing tool is registered directly by CPE and does not require MCP configuration. It is enabled by default; set disable_edit_tool: true on a model profile to omit it.

```yaml

Ask a question (or set CPE_MODEL=sonnet once in your shell)

cpe --model sonnet "Explain what this project does"

🔧 Troubleshooting