CLI Tool Connectors

DgiDgi supports integration with CLI-based AI coding assistants. These tools run as terminal subprocesses and provide interactive AI assistance directly in the command line.

Supported CLI Tools

Anthropic Claude Code

The official CLI for Claude - agentic coding assistant that can edit files, run commands, and work autonomously.

Feature	Support
Interactive Chat	Yes
Code Generation	Yes
File Editing	Yes
Git Operations	Yes
Terminal Access	Yes
Agentic Mode	Yes

Installation:

npm install -g @anthropic-ai/claude-code
export ANTHROPIC_API_KEY=sk-ant-...
claude --help

Usage in DgiDgi:

// Via CLI executor
await executeCLITool("cli.invoke", {
  provider: "claude-code",
  prompt: "Add error handling to the API routes",
  files: ["server/routes/api.ts"],
});

---

GitHub Copilot CLI

GitHub Copilot in the command line via the gh extension.

Feature	Support
Interactive Chat	Yes
Code Generation	Yes
File Editing	No
Git Operations	Yes
Terminal Access	Yes
Agentic Mode	No

Installation:

gh extension install github/gh-copilot
gh auth login
gh copilot suggest "how to create a new branch"

---

Aider

Open-source AI pair programming that works with multiple LLM providers.

Feature	Support
Interactive Chat	Yes
Code Generation	Yes
File Editing	Yes
Git Operations	Yes
Terminal Access	No
Agentic Mode	Yes

Installation:

pip install aider-chat
export ANTHROPIC_API_KEY=sk-ant-...  # or OPENAI_API_KEY
aider --help

Features:

• Works with Claude, GPT-4, local models
• Automatic git commits for changes
• Voice input support
• Repository map for context

---

Sourcegraph Cody CLI

Sourcegraph's AI coding assistant with code intelligence.

Feature	Support
Interactive Chat	Yes
Code Generation	Yes
File Editing	Yes
Git Operations	No
Terminal Access	No
Agentic Mode	No

Installation:

npm install -g @sourcegraph/cody
export SRC_ACCESS_TOKEN=...
cody chat "explain this function"

---

Amazon Q Developer

AWS AI-powered developer assistant with AWS integration.

Feature	Support
Interactive Chat	Yes
Code Generation	Yes
File Editing	Yes
Git Operations	No
Terminal Access	Yes
Agentic Mode	Yes

Installation:

brew install amazon-q
# or download from AWS
q developer login
q chat "help me deploy to Lambda"

---

Continue

Open-source AI code assistant with VS Code and JetBrains extensions.

Feature	Support
Interactive Chat	Yes
Code Generation	Yes
File Editing	Yes
Git Operations	No
Terminal Access	No
Agentic Mode	No

Installation:

npm install -g @continue/cli
continue config  # Configure your LLM provider
continue ask "refactor this code"

---

Cursor CLI

Cursor AI editor's command line interface.

Feature	Support
Interactive Chat	Yes
Code Generation	Yes
File Editing	Yes
Git Operations	No
Terminal Access	Yes
Agentic Mode	Yes

Installation: Download Cursor from cursor.com

---

DgiDgi CLI

DgiDgi's own CLI tool for AI-powered development.

Feature	Support
Interactive Chat	Yes
Code Generation	Yes
File Editing	Yes
Git Operations	Yes
Terminal Access	Yes
Agentic Mode	Yes

Installation:

pnpm add -g @dgidgi-one/cli
dgidgi login
dgidgi ask "help me build a new feature"

Features:

• Connects to your DgiDgi projects
• Uses your configured LLM providers
• Syncs with WebContainer runtime
• Full agentic task execution

Provider Scopes

DgiDgi supports two provider scopes:

Scope	Description	Billing
Tenant	Configured per organization/tenant	Tenant pays
CLI	Terminal-based AI tools running as subprocesses	User pays (own keys)

API Routes

List CLI Providers

GET /api/v1/cli-providers

Get Provider Details

GET /api/v1/cli-providers/:id

Check Installation

POST /api/v1/cli-providers/:id/check
{
  "workspacePath": "/path/to/project"
}

Install Provider

POST /api/v1/cli-providers/:id/install

Filter by Capability

GET /api/v1/cli-providers/by-capability/agenticMode

Using CLI Tools in Agents

The DgiDgi agent system can invoke CLI tools:

// List available CLI tools
const tools = await executeCLITool("cli.list", {}, context, events, deps);

// Check if a tool is installed
const status = await executeCLITool("cli.check", {
  provider: "claude-code"
}, context, events, deps);

// Invoke a CLI tool
const result = await executeCLITool("cli.invoke", {
  provider: "aider",
  prompt: "Add unit tests for the user service",
  files: ["src/services/user.service.ts"],
  timeout: 120000,
}, context, events, deps);

// Spawn interactive session
const session = await executeCLITool("cli.spawn", {
  provider: "claude-code"
}, context, events, deps);

// Send input to session
await executeCLITool("cli.send", {
  processId: session.output.processId,
  input: "explain this codebase"
}, context, events, deps);

// Kill session
await executeCLITool("cli.kill", {
  processId: session.output.processId
}, context, events, deps);

Security Considerations

CLI tools run as subprocesses with the following security measures:

1. Workspace Isolation: Commands run within the project workspace
2. Secret Redaction: Output is scanned and secrets are redacted
3. Timeout Enforcement: Commands have configurable timeouts (default 2 min)
4. Process Tracking: All spawned processes are tracked and can be killed
5. Environment Isolation: Custom env vars can override process.env

Comparison Table

Tool	Underlying LLM	Pricing	Best For
Claude Code	Claude	Pay-per-token	Full agentic coding
Copilot CLI	GPT-4	Subscription	Quick suggestions
Aider	Any (configurable)	Free (bring your key)	Open-source flexibility
Cody CLI	Claude/GPT	Freemium	Codebase understanding
Amazon Q	Bedrock models	Freemium	AWS development
Continue	Any (configurable)	Free	IDE integration
Cursor	Custom	Freemium	Editor-native AI
DgiDgi CLI	DgiDgi providers	Freemium	Platform integration