feat: add AI Agent integration with MCP Server and CLI structured output

[lxcxjxhx]

Summary

This PR adds comprehensive AI Agent integration to Sliver C2, enabling modern AI-powered automation and orchestration capabilities.

Changes

1. MCP Server Implementation

• Implemented Model Context Protocol (MCP) Server with 20+ tools for AI Agent interaction
• Tools include: session/beacon management, file operations, process management, credential handling, network reconnaissance
• Supports both stdio and HTTP transport layers
• API Token authentication for secure access

2. CLI Structured Output

• Added --output-format flag supporting JSON and YAML formats
• Shorthand flags: --json and --yaml for convenience
• Applied to key commands: sessions, beacons, loot, creds, etc.
• Enables easy parsing by AI Agents and automation scripts

3. Documentation

• Comprehensive AI Agent integration guide in docs/ai-agent-integration.md
• Usage examples for MCP Server and CLI structured output
• Best practices for AI Agent integration

Motivation

Modern penetration testing and red team operations increasingly leverage AI Agents for automation, decision-making, and orchestration. This integration enables:

• AI-Powered Automation: AI Agents can directly interact with Sliver through standardized MCP protocol
• Structured Data Exchange: JSON/YAML output allows easy parsing and analysis
• Tool Integration: MCP tools provide granular control over Sliver operations
• Future-Proof Architecture: MCP is becoming the standard for AI tool integration

Testing

• Local compilation tested (Go 1.21+)
• MCP Server tested with MCP Inspector
• CLI structured output verified across multiple commands
• All existing tests pass

Files Changed

• client/mcp/: New MCP Server implementation (server.go, transport.go, auth.go, tools)
• client/command/output/: Structured output support
• docs/ai-agent-integration.md: Integration documentation
• Modified CLI commands to support structured output

Example Usage

MCP Server

sliver mcp-server --transport stdio

CLI Structured Output

sliver sessions --json
sliver beacons --yaml
sliver loot list --output-format json

This PR significantly enhances Sliver's automation capabilities while maintaining backward compatibility with existing workflows.

[lxcxjxhx]

Thank you for your contribution! However, after reviewing the PR, I found several issues:

1. MCP Support Already Exists: The project already has MCP infrastructure merged (Experiamental MCP Support #2083, Fix/mcp #2228), so this isn't adding new AI Agent integration from scratch.

2. Duplicate Work: PR feat(mcp): add execute tool for remote command execution #2277 (opened 20 days earlier on June 17) is already adding the execute tool, which overlaps with the client/mcp/execute.go in this PR.

3. Scope Concerns: This PR adds 20+ MCP tools at once (credentials, network, pivot, processes, systeminfo, upload, etc.) plus CLI structured output. The project appears to prefer incremental additions rather than bulk tool deployments.

Recommendations:

• For the execute tool: Please follow and contribute to feat(mcp): add execute tool for remote command execution #2277 instead
• For other MCP tools: Consider splitting into multiple smaller, focused PRs that can be reviewed individually
• For CLI structured output (JSON/YAML): This could be a separate PR as it's a distinct feature

This approach aligns better with the project's development style and allows maintainers to properly review each addition. Thank you for understanding!