Google Vertex AI Agent with MCP Security Tools

Warning

One user has reported ~$30/day expense in Spanner and a concern that it is due to the RAG Corpus from this project. I do not see this in my own projects but I am investigating further. In the meantime, please carefully monitor your expenses.

Google Vertex AI Agent with MCP Security Tools

Deploy security-focused AI agents to Google Cloud with integrated access to Chronicle, SOAR, Threat Intelligence, and Security Command Center through the Model Context Protocol (MCP).

Quick Start
Overview
Architecture
Prerequisites
Installation
Deployment Workflow
Configuration
Usage
AgentSpace Integration
Makefile Reference
Python CLI (Alternative Interface)
Project Structure
Development
- Local Development with ADK Web
- Testing Individual MCP Servers
Troubleshooting
FAQ
Best Practices
Documentation
Support

Quick Start

Option 1: Local Development (Recommended for Getting Started)

# Clone and setup
git clone --recurse-submodules https://github.com/dandye/agentic_soc_agentspace.git
cd agentic_soc_agentspace

# Configure environment
cp .env.example .env
# Edit .env with your Google Cloud credentials

# Install dependencies in a virtual env
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Run locally with ADK Web (instant, no deployment needed!)
cd soc_agent
GOOGLE_GENAI_USE_VERTEXAI=True \
GOOGLE_APPLICATION_CREDENTIALS=/path/to/your/service-account.json \
GOOGLE_CLOUD_PROJECT=your-project-id \
GOOGLE_CLOUD_LOCATION=us-central1 \
adk web

This opens an interactive web UI at http://localhost:8000 where you can test all features instantly. See Local Development with ADK Web for details.

Option 2: Production Deployment to Agent Engine

# After completing the setup above:

# 1. Verify Vertex AI setup (one-time verification)
python manage.py vertex verify

# 2. Configure IAM permissions (one-time setup)
python manage.py iam setup

# 3. Deploy to Agent Engine
make agent-engine-deploy

For detailed deployment instructions, see the Deployment Workflow section below.

Available Commands

The Makefile provides a comprehensive set of commands for managing your deployment:

Run make help to see this interactive command reference.

Overview

This project enables you to:

Deploy AI agents to Google Vertex AI Agent Engine with security tool access
Integrate with Google Security Operations (Chronicle) for threat detection
Connect to SOAR platforms for automated response workflows
Access Google Threat Intelligence for IOC analysis
Monitor cloud security posture via Security Command Center

Architecture

┌─────────────────┐    ┌──────────────────┐    ┌─────────────────────┐
│   main.py       │───▶│  Vertex AI       │───▶│   MCP Security      │
│   Agent Creator │    │  Agent Engine    │    │   Tools (stdio)     │
└─────────────────┘    └──────────────────┘    └─────────────────────┘
                               │                        │
┌─────────────────┐            │                ┌───────▼───────┐
│test_agent_engine│◀───────────┘                │ • Chronicle   │
│    Testing      │                             │ • SOAR        │
└─────────────────┘                             │ • GTI         │
                                                │ • SCC         │
                                                └───────────────┘

Prerequisites

System Requirements

Python 3.10+ installed locally
4GB RAM minimum
Git with submodules support
Google Cloud SDK installed and configured

Google Cloud Requirements

Active GCP project with billing enabled
Project ID and Project Number available

Required APIs

Enable these APIs in your project:

gcloud services enable aiplatform.googleapis.com
gcloud services enable storage.googleapis.com
gcloud services enable cloudbuild.googleapis.com
gcloud services enable compute.googleapis.com
gcloud services enable discoveryengine.googleapis.com
gcloud services enable securitycenter.googleapis.com  # If using SCC tools

Required IAM Permissions

Your account needs these roles:

roles/aiplatform.user - Create and manage AI Platform resources
roles/storage.admin - Manage staging bucket
roles/iam.serviceAccountUser - Create service accounts for agent
roles/discoveryengine.admin - Manage AgentSpace and register agents (includes agents.manage permission)
roles/securitycenter.admin - Access Security Command Center (if using)

IMPORTANT: The roles/discoveryengine.admin role is required to register agents with AgentSpace (provides the agents.manage permission). Regular users can use registered agents but cannot register new ones.

Service Account Configuration

Quick Setup (Recommended):

Use the IAM management CLI to configure all required service account permissions automatically:

# Setup all required IAM permissions for AgentSpace
python manage.py iam setup

# Verify permissions are configured correctly
python manage.py iam verify

# Preview changes before applying (optional)
python manage.py iam setup --dry-run --verbose

What this configures:

AI Platform Reasoning Engine Service Agent
- Service Account: service-{PROJECT_NUMBER}@gcp-sa-aiplatform-re.iam.gserviceaccount.com
- Role: roles/aiplatform.user
- Purpose: Query RAG corpus during agent execution
- Fixes: 403 PERMISSION_DENIED error for aiplatform.ragCorpora.query
Discovery Engine Service Account
- Service Account: service-{PROJECT_NUMBER}@gcp-sa-discoveryengine.iam.gserviceaccount.com
- Roles: roles/aiplatform.user, roles/aiplatform.viewer
- Purpose: Call ADK agent from AgentSpace
- Fixes: Agent registered but cannot be invoked from Gemini Enterprise UI

Manual Setup (Alternative):

If you prefer to configure permissions manually using gcloud:

# Get your project number
GCP_PROJECT_NUMBER=$(gcloud projects describe $GCP_PROJECT_ID --format="value(projectNumber)")

# 1. AI Platform Reasoning Engine Service Agent (for RAG access)
gcloud projects add-iam-policy-binding $GCP_PROJECT_ID \
  --member="serviceAccount:service-${GCP_PROJECT_NUMBER}@gcp-sa-aiplatform-re.iam.gserviceaccount.com" \
  --role="roles/aiplatform.user"

# 2. Discovery Engine Service Account (for AgentSpace integration)
gcloud projects add-iam-policy-binding $GCP_PROJECT_ID \
  --member="serviceAccount:service-${GCP_PROJECT_NUMBER}@gcp-sa-discoveryengine.iam.gserviceaccount.com" \
  --role="roles/aiplatform.user"

gcloud projects add-iam-policy-binding $GCP_PROJECT_ID \
  --member="serviceAccount:service-${GCP_PROJECT_NUMBER}@gcp-sa-discoveryengine.iam.gserviceaccount.com" \
  --role="roles/aiplatform.viewer"

Verification:

You can verify the configuration in the IAM Console by checking "Include Google-provided role grants" to view Google-managed service accounts, or use the CLI:

# List all roles for a service account
python manage.py iam list-roles aiplatform-re
python manage.py iam list-roles discoveryengine

See Google Cloud IAM Service Agents documentation for more details about Google-managed service accounts.

Authentication Setup

GCP_PROJECT_ID=your-project-id
gcloud auth application-default login
gcloud config set project $GCP_PROJECT_ID
gcloud auth application-default set-quota-project $GCP_PROJECT_ID

Installation

1. Clone Repository (note that the submodule is checked out too)

git clone --recurse-submodules https://github.com/dandye/agentic_soc_agentspace.git
cd agentic_soc_agentspace

2. Configure Environment

cp .env.example .env
# Edit .env with your configuration
# NOTE: Some vars, like `AGENT_ENGINE_RESOURCE_NAME`, are generated by make targets (i.e. `make agent-engine-deploy`)
#  so they are not available to you yet.

3. Install Dependencies

pip install -r requirements.txt

4. Deploy Agent

make agent-engine-deploy
# Or directly: python main.py

5. Update .env with deployment outputs

Set AGENT_ENGINE_RESOURCE_NAME in .env (output from deployment)

6. Create AgentSpace App (Optional)

Option A: Using the Web UI (Recommended)

Navigate to Google Cloud Console
Go to Vertex AI > Search & Conversation > Apps
Click Create App
Select Agent as the app type
Configure with your preferred settings
Copy the App ID and add to .env as AGENTSPACE_APP_ID

Option B: Using the CLI (Advanced)

Important

CRITICAL: When creating apps via the API, you MUST include --app-type APP_TYPE_INTRANET and --industry-vertical GENERIC for apps to appear in the Gemini Enterprise web UI. Without these fields, apps may be created successfully via API but will not be visible in the console UI.

See the official Google Cloud documentation for details.

# Create app with proper visibility settings
python manage.py agentspace create-app \
  --name "My Security Agent" \
  --type SOLUTION_TYPE_CHAT \
  --no-datastore \
  --app-type APP_TYPE_INTRANET \
  --industry-vertical GENERIC

7. Register Agent with AgentSpace

make agentspace-register

Deployment Workflow

The deployment process consists of four stages:

Stage 0: Prerequisites Verification and Setup (One-time Setup)

Before deploying, verify and configure your Google Cloud environment:

Step 1: Verify Vertex AI Setup

Run the comprehensive verification check:

python manage.py vertex verify

This verifies:

Environment variables are set correctly
Application Default Credentials are configured
GCP project is accessible
Required APIs are enabled
Vertex AI can initialize
Your user account has required IAM roles

If checks fail:

# Fix authentication
gcloud auth application-default login
gcloud auth application-default set-quota-project PROJECT_ID

# Enable required APIs
python manage.py vertex enable-apis

# Or manually:
gcloud services enable aiplatform.googleapis.com storage.googleapis.com \
  cloudbuild.googleapis.com compute.googleapis.com discoveryengine.googleapis.com

Step 2: Configure IAM Service Account Permissions

Configure required service account permissions for AgentSpace integration:

# Setup IAM permissions (run once per project)
python manage.py iam setup

# Verify configuration
python manage.py iam verify

This configures:

AI Platform Reasoning Engine Service Agent (for RAG access)
Discovery Engine Service Account (for calling agents from AgentSpace)

See Required IAM Permissions for details.

Stage 1: Prerequisites (User Configuration)

Set these variables in your .env file before deployment:

Required Core Variables:

GCP_PROJECT_ID - Your Google Cloud Project ID
GCP_PROJECT_NUMBER - Your Google Cloud Project Number (numeric ID)
GCP_LOCATION - Deployment region (e.g., us-central1)
GCP_STAGING_BUCKET - GCS bucket name with gs:// prefix (e.g., gs://my-bucket)

Required Security Tools:

Chronicle SIEM: CHRONICLE_PROJECT_ID, CHRONICLE_CUSTOMER_ID, CHRONICLE_SERVICE_ACCOUNT_PATH
SOAR: SOAR_URL, SOAR_API_KEY
Threat Intelligence: GTI_API_KEY (Google Threat Intelligence/VirusTotal)
RAG Corpus: RAG_CORPUS_ID (full resource name)

Optional Variables (have sensible defaults):

CHRONICLE_REGION - Chronicle region (default: "us")
DEBUG - Enable debug logging (default: False)
RAG_SIMILARITY_TOP_K - RAG retrieval top-k results (default: 10)
RAG_DISTANCE_THRESHOLD - RAG vector distance threshold (default: 0.6)

Stage 2: Deployment Outputs

Run make agent-engine-deploy and save the generated AGENT_ENGINE_RESOURCE_NAME from the output to your .env.

Stage 3: Integration Outputs

Run make agentspace-register and save the generated AGENTSPACE_AGENT_ID to your .env.

Configuration

See .env.example for all environment variables with comprehensive documentation. Key variables are documented in the Deployment Workflow section above.

For migration from v1.x to v2.0, see BREAKING_CHANGES.md.

Usage

Deploy New Agent

# Full deployment with all tools
make agent-engine-deploy

Test Deployed Agent

# After setting AGENT_ENGINE_RESOURCE_NAME in .env from deployment output
python test_agent_engine.py

Example Queries

"List soar cases"
"What are the recent security alerts?"
"Analyze IOCs for domain malicious.com"
"Show Security Command Center findings"
"Investigate user account compromise indicators"
"Find a runbook titled Malware IRP"

Workflow Commands

make full-deploy - Complete standard deployment
make full-deploy-with-oauth - Deployment with OAuth
make redeploy-all - Redeploy and update AgentSpace

RAG Corpus Management

Create and manage RAG (Retrieval-Augmented Generation) corpora for document search:

# List all RAG corpora
make rag-list

# List with detailed information
make rag-list VERBOSE=1

# Create a new RAG corpus
make rag-create NAME="Security Runbooks"

# Get corpus information
make rag-info RAG_CORPUS_ID=projects/PROJECT/locations/LOCATION/ragCorpora/CORPUS_ID

# Delete a corpus
make rag-delete RAG_CORPUS_ID=projects/PROJECT/locations/LOCATION/ragCorpora/CORPUS_ID

Note: After creating a RAG corpus, save the resource name to your .env as RAG_CORPUS_ID.

Data Store Management (Optional - Legacy)

Discovery Engine data stores are available for AgentSpace apps but RAG is the recommended approach:

# List data stores
make datastore-list

# Create a data store
make datastore-create NAME="Security Data" TYPE=SOLUTION_TYPE_SEARCH

AgentSpace Integration

Important

APP_TYPE_INTRANET Required: When creating apps programmatically via the API/CLI, you MUST include appType=APP_TYPE_INTRANET and industryVertical=GENERIC for apps to be visible in the Gemini Enterprise web UI. Apps created through the console UI include these fields automatically. See the official documentation.

Option 1: Create via Console (Recommended)

Create AgentSpace App - Via Console
- Navigate to Vertex AI > Search & Conversation > Apps
- Click Create App and select Agent type
- Configure with your preferred settings
- Note: The agent uses RAG for document retrieval, not Discovery Engine data stores
Copy App ID to .env as AGENTSPACE_APP_ID

Option 2: Create via CLI (Advanced)

python manage.py agentspace create-app \
  --name "Security Agent" \
  --type SOLUTION_TYPE_CHAT \
  --no-datastore \
  --app-type APP_TYPE_INTRANET \
  --industry-vertical GENERIC

After creating the app: 3. Register agent: make agentspace-link-agent (add OAuth with make oauth-setup first if needed) 4. Verify: make agentspace-verify

Makefile Commands

Run make help to see all available commands with descriptions (shown in image above).

Key Environment Flags

FORCE=1 - Skip confirmation prompts
V=1 - Verbose output
INDEX=n - Specify item index

Example: FORCE=1 make agentspace-register

Python CLI (Alternative Interface)

In addition to the Makefile, this project provides a unified Python CLI built with Typer that offers the same functionality with additional benefits like type safety, autocomplete, and programmatic access.

Quick Start

# Show all available commands
python manage.py --help

# Check system status
python manage.py workflow status

# List agent engines
python manage.py agent-engine list

# Register with AgentSpace
python manage.py agentspace register

CLI Structure

The CLI is organized into subcommand groups:

python manage.py
├── agent-engine    # Manage Agent Engine instances
├── agentspace      # Manage AgentSpace apps and agents
├── vertex          # Verify and manage Vertex AI setup
├── iam             # Manage IAM permissions for service accounts
├── oauth           # Manage OAuth authorizations
├── datastore       # Manage data stores
├── rag             # Manage RAG corpora
├── workflow        # Composite workflows
├── setup           # Environment setup
└── version         # Version information

Common Commands

Agent Engine Management:

python manage.py agent-engine list
python manage.py agent-engine delete --index 1
python manage.py agent-engine delete --resource "projects/.../reasoningEngines/..." --force

Vertex AI Setup:

python manage.py vertex verify              # Verify complete setup
python manage.py vertex verify --skip-apis  # Skip API checks
python manage.py vertex enable-apis         # Enable required APIs
python manage.py vertex check-quota         # Display quota information

AgentSpace Management:

python manage.py agentspace register
python manage.py agentspace update
python manage.py agentspace verify
python manage.py agentspace link-agent
python manage.py agentspace list-agents

IAM Permissions Management:

python manage.py iam setup                    # Configure all required permissions
python manage.py iam setup --dry-run          # Preview changes without applying
python manage.py iam setup --verbose          # Show detailed information
python manage.py iam verify                   # Verify all permissions are configured
python manage.py iam list-roles aiplatform-re # List roles for AI Platform RE
python manage.py iam list-roles discoveryengine # List roles for Discovery Engine

RAG Corpus Management:

python manage.py rag list --verbose
python manage.py rag create "Security Runbooks" --description "SOC procedures"
python manage.py rag info "projects/.../ragCorpora/..."
python manage.py rag delete "projects/.../ragCorpora/..." --force

OAuth Management:

python manage.py oauth setup client_secret.json
python manage.py oauth create-auth
python manage.py oauth verify

Workflow Commands:

# Complete deployment with OAuth
python manage.py workflow full-deploy

# Redeploy everything
python manage.py workflow redeploy-all

# Check system status
python manage.py workflow status

Command Mapping: Makefile to Python CLI

Makefile Command	Python CLI Equivalent
`make agent-engine-list`	`python manage.py agent-engine list`
`make agent-engine-delete-by-index INDEX=1`	`python manage.py agent-engine delete --index 1`
`make agentspace-register`	`python manage.py agentspace register`
`make agentspace-register FORCE=1`	`python manage.py agentspace register --force`
`make oauth-setup CLIENT_SECRET=x.json`	`python manage.py oauth setup x.json`
`make rag-list`	`python manage.py rag list`
`make rag-create NAME="x"`	`python manage.py rag create "x"`
`make full-deploy-with-oauth`	`python manage.py workflow full-deploy`
`make status`	`python manage.py workflow status`

Benefits Over Makefile

Type Safety: Typer provides parameter validation and type checking
Autocomplete: Shell completion for all commands and options
Better Help: Rich formatted help with detailed descriptions
Cross-Platform: Works on all platforms without Make dependency
Programmatic Use: Can be imported and used as a Python library
Consistent API: All commands follow the same pattern

Environment File Support

All commands support custom environment files:

python manage.py --env-file .env.prod agent-engine list
python manage.py --env-file .env.staging workflow status

Shell Autocomplete

Install autocomplete for your shell:

# Bash
python manage.py --install-completion bash

# Zsh
python manage.py --install-completion zsh

# Fish
python manage.py --install-completion fish

Detailed Documentation

For comprehensive CLI documentation including all commands, options, and examples, see MANAGE_CLI_USAGE.md.

Project Structure

├── main.py                    # Agent creation and deployment
├── test_agent_engine.py       # Testing interface
├── DEPLOYMENT_GUIDE.md        # Detailed deployment instructions
├── requirements.txt           # Python dependencies
├── Makefile                   # Automation workflows
├── .env.example               # Environment template with documentation
├── .gitmodules                # Git submodule configuration
├── installation_scripts/
│   ├── install.sh             # Cloud deployment setup (installs `uv`)
│   ├── manage_agent_engine.py # Agent management utilities
│   ├── manage_agentspace.py   # Agentspace configuration
│   └── manage_oauth.py        # OAuth setup utilities
└── mcp-security/              # MCP security servers (submodule)
    └── server/
        ├── mcp-server-soar/
        ├── mcp-server-secops/
        ├── mcp-server-gti/
        └── mcp-server-scc/

Development

Code Quality and Testing

This project uses a modern code quality stack to maintain security-critical code:

Install development tools:

pip install -r requirements-dev.txt

Pre-commit hooks (automatic on git commit):

pre-commit install              # Install hooks
pre-commit run --all-files      # Run manually

Run code quality checks:

ruff check .                    # Linting and security checks
ruff check --fix .              # Auto-fix issues
pyink .                         # Format code (Google style)
mypy .                          # Type checking
pytest                          # Run tests with coverage
pip-audit                       # Dependency vulnerability scan

Tools used:

pyink: Code formatting (Google Python Style Guide)
ruff: Fast linting, import sorting, security checks
mypy: Static type checking
pytest: Testing with coverage reporting
pip-audit: Dependency vulnerability scanning

CI/CD: All PRs run automated checks for linting, type checking, security scanning, and test coverage.

Customizing the Agent

Edit main.py to customize:

# Change model
model = "gemini-2.5-flash"  # or gemini-2.5-pro

# Modify system prompt
instruction = "Your custom security analyst instructions..."

# Select specific tools
tools = [secops_tool, soar_tool]  # Choose tools as needed

Adding New MCP Tools

Add tool configuration to MCPToolset in main.py
Update extra_packages with server path
Modify installation script if needed

Local Development with ADK Web

The fastest way to develop and test your agent locally is using adk web, which provides an interactive web UI without deploying to Agent Engine.

Running ADK Web

From the soc_agent directory, run:

cd soc_agent

GOOGLE_GENAI_USE_VERTEXAI=True \
GOOGLE_APPLICATION_CREDENTIALS=/path/to/your/service-account.json \
GOOGLE_CLOUD_PROJECT=your-project-id \
GOOGLE_CLOUD_LOCATION=us-central1 \
adk web

Replace the values:

/path/to/your/service-account.json - Path to your Chronicle service account JSON file (same as CHRONICLE_SERVICE_ACCOUNT_PATH in .env)
your-project-id - Your GCP project ID (same as GCP_PROJECT_ID in .env)
us-central1 - Your GCP location (same as GCP_LOCATION in .env)

The web UI will open at http://localhost:8000 where you can:

Chat with your agent in real-time
Test MCP tool integrations (Chronicle, SOAR, GTI, SCC)
Query the RAG corpus for runbooks and procedures
See tool calls and responses as they happen
Iterate on your agent configuration without redeployment

Why Use ADK Web?

Benefits:

Instant feedback: No 10-15 minute deployment wait time
Cost effective: No Agent Engine hosting costs during development
Full feature parity: All MCP tools and RAG retrieval work locally
Easy debugging: See tool calls and responses in real-time
Rapid iteration: Modify agent.py and restart immediately

When to use Agent Engine instead:

Production deployments
Integration with AgentSpace
Multi-user access
Persistent conversation history
Autoscaling requirements

Testing Individual MCP Servers

You can also test MCP servers in isolation:

# Test MCP servers locally
cd mcp-security/server/secops-soar/secops_soar_mcp
uv run server.py

Success looks like:

2025-09-15 16:39:39,549 - INFO - __main__ - main - Starting SecOps SOAR MCP server
2025-09-15 16:39:39,817 - INFO - __main__ - get_enabled_integrations_set - No --integrations flag provided. No integrations are enabled.
2025-09-15 16:39:39,818 - INFO - __main__ - register_tools - Starting dynamic tool registration...
2025-09-15 16:39:39,824 - INFO - __main__ - register_tools - Finished scanning marketplace directory.

(and then hanging as it is waiting for input)

Validate configuration

python -c "from dotenv import load_dotenv; import os; load_dotenv(); print(os.getenv('GCP_PROJECT_ID'))"

Troubleshooting

Common Issues & Quick Fixes

Issue	Solution
403/401 Auth Error	`gcloud auth application-default login`
API not enabled	`gcloud services enable aiplatform.googleapis.com`
Bucket not found	`gsutil mb -p $GCP_PROJECT_ID $GCP_STAGING_BUCKET`
MCP module missing	`git submodule update --init --recursive`
KeyError: GCP_PROJECT_ID	Check `.env` exists with correct variable names (v2.0+)
Agent not in AgentSpace	`make agentspace-verify` then `make agentspace-link-agent`
OAuth expired	`make oauth-setup CLIENT_SECRET=client_secret.json`
Agent not responding	Check logs: `gcloud logging tail "resource.type=aiplatform.googleapis.com/ReasoningEngine"`
SOAR connection failed	Verify URL: `curl -I "${SOAR_URL}/api/external/v1/health"`

Quick Debug Commands

gcloud logging tail "resource.type=aiplatform.googleapis.com/ReasoningEngine"  # Watch logs
make agentspace-test  # Test components

FAQ

Can I use this without SOAR? Yes. All security tool integrations are optional.

What AI models are supported? The default is gemini-2.5-flash but you can use any model suppored by Vertex AI Model Garden and ADK. Configure in main.py.

How do I update the agent?

git pull && make agent-engine-deploy && make agentspace-update

What are the costs? Vertex AI charges per API call. Security products require separate licensing.

Best Practices

Security: Use Secret Manager for credentials, enable audit logging, apply least-privilege IAM
Development: Separate dev/staging/prod projects, version control all configuration
Operations: Set budget alerts, backup configurations, monitor quotas

Documentation

MCP Security Docs - Security tool documentation
Google Vertex AI Docs - Platform documentation
MCP Protocol - Protocol specification

Support

Getting Help

GitHub Issues - Report bugs or request features
Stack Overflow - Community support
Google Cloud Support - Production issues

Resources

Vertex AI Docs - Google Cloud documentation
MCP Protocol - Model Context Protocol specification
Chronicle Docs - SIEM documentation
Security Command Center - Cloud security docs

Name		Name	Last commit message	Last commit date
Latest commit History 134 Commits
.github/workflows		.github/workflows
adk_runbooks @ 6465a29		adk_runbooks @ 6465a29
ai-runbooks @ fca5f8e		ai-runbooks @ fca5f8e
installation_scripts		installation_scripts
mcp-security @ 1526489		mcp-security @ 1526489
soc_agent		soc_agent
soc_agent_cti		soc_agent_cti
soc_agent_flash		soc_agent_flash
soc_agent_tier1		soc_agent_tier1
.env.example		.env.example
.gitignore		.gitignore
.gitmodules		.gitmodules
.pre-commit-config.yaml		.pre-commit-config.yaml
GEMINI.md		GEMINI.md
Makefile		Makefile
README.md		README.md
manage.py		manage.py
pyproject.toml		pyproject.toml
requirements-dev.txt		requirements-dev.txt
requirements.in		requirements.in
requirements.txt		requirements.txt

dandye/agentic_soc_agentspace

Folders and files

Latest commit

History

Repository files navigation

Google Vertex AI Agent with MCP Security Tools

Table of Contents

Quick Start

Option 1: Local Development (Recommended for Getting Started)

Option 2: Production Deployment to Agent Engine

Available Commands

Overview

Architecture

Prerequisites

System Requirements

Google Cloud Requirements

Required APIs

Required IAM Permissions

Service Account Configuration

Authentication Setup

Installation

1. Clone Repository (note that the submodule is checked out too)

2. Configure Environment

3. Install Dependencies

4. Deploy Agent

5. Update .env with deployment outputs

6. Create AgentSpace App (Optional)

7. Register Agent with AgentSpace

Deployment Workflow

Stage 0: Prerequisites Verification and Setup (One-time Setup)

Step 1: Verify Vertex AI Setup

Step 2: Configure IAM Service Account Permissions

Stage 1: Prerequisites (User Configuration)

Stage 2: Deployment Outputs

Stage 3: Integration Outputs

Configuration

Usage

Deploy New Agent

Test Deployed Agent

Example Queries

Workflow Commands

RAG Corpus Management

Data Store Management (Optional - Legacy)

AgentSpace Integration

Makefile Commands

Key Environment Flags

Python CLI (Alternative Interface)

Quick Start

CLI Structure

Common Commands

Command Mapping: Makefile to Python CLI

Benefits Over Makefile

Environment File Support

Shell Autocomplete

Detailed Documentation

Project Structure

Development

Code Quality and Testing

Customizing the Agent

Adding New MCP Tools

Local Development with ADK Web

Running ADK Web

Why Use ADK Web?

Testing Individual MCP Servers

Validate configuration

Troubleshooting

Common Issues & Quick Fixes

Quick Debug Commands

FAQ

Best Practices

Documentation

Support

Getting Help

Resources

About

Resources

Uh oh!

Stars

Watchers

Forks

Packages