### Download setup script and example config

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/aws?q=

Downloads the setup script and example configuration file for the multi-account AWS MCP setup. Run these commands before proceeding to the configuration step.

```bash
curl -O https://raw.githubusercontent.com/robusta-dev/holmes-mcp-integrations/master/servers/aws/setup-multi-account-iam.sh
chmod +x setup-multi-account-iam.sh

# Download example configuration file
curl -O https://raw.githubusercontent.com/robusta-dev/holmes-mcp-integrations/master/servers/aws/multi-cluster-config-example.yaml

```

--------------------------------

### Download multi-account setup script and example config

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/aws

Downloads the multi-account setup script and an example configuration file from the Holmes MCP integrations repository. Make the script executable before running it.

```bash
# Download the setup script
curl -O https://raw.githubusercontent.com/robusta-dev/holmes-mcp-integrations/master/servers/aws/setup-multi-account-iam.sh
chmod +x setup-multi-account-iam.sh

# Download example configuration file
curl -O https://raw.githubusercontent.com/robusta-dev/holmes-mcp-integrations/master/servers/aws/multi-cluster-config-example.yaml

```

--------------------------------

### Quick start with AWS Bedrock

Source: https://holmesgpt.dev/latest/installation/cli-installation?q=

Quick start with AWS Bedrock. Set up AWS credentials, install boto3, create a test pod, and ask your first question using a Bedrock model. See AWS Bedrock Configuration for more details.

```bash
export AWS_ACCESS_KEY_ID="your-access-key"
export AWS_SECRET_ACCESS_KEY="your-secret-key"
export AWS_DEFAULT_REGION="your-region"

```

```bash
pip install "boto3>=1.28.57"

```

```bash
kubectl apply -f https://raw.githubusercontent.com/robusta-dev/kubernetes-demos/main/pending_pods/pending_pod_node_selector.yaml

```

```bash
holmes ask "what is wrong with the user-profile-import pod?" --model="bedrock/us.anthropic.claude-sonnet-4-5-20250929-v1:0"

# Or use another model
holmes ask "what is wrong with the user-profile-import pod?" --model="bedrock/<your-model-name>"

```

--------------------------------

### Install and start HolmesGPT HTTP server with Docker Compose

Source: https://holmesgpt.dev/latest/installation/docker-compose-installation

Steps to install and start the HolmesGPT HTTP server with Docker Compose: clone the repository, set the OpenAI API key, start the server, and verify it is running.

```bash
git clone https://github.com/HolmesGPT/holmesgpt.git
cd holmesgpt

```

```bash
export OPENAI_API_KEY="your-api-key"

```

```bash
docker compose up

```

```bash
curl http://localhost:5050/healthz

```

--------------------------------

### Example usage: list tables and sizes

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-postgresql

Prompt-style example asking for a list of all tables and their sizes.

```text
"List all tables and their sizes"
```

--------------------------------

### Example usage: get table schema

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-postgresql

Prompt-style example asking HolmesGPT for the schema of the users table.

```text
"Show me the schema for the users table"
```

--------------------------------

### List tables prompt

Source: https://holmesgpt.dev/latest/data-sources/database-connectors?q=

Example prompt asking Holmes to list all tables and their sizes in the prod-postgres toolset. Use this to get an overview of the database.

```text
"List all tables and their sizes in prod-postgres"

```

--------------------------------

### Run multi-account setup script

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/aws?q=

Commands to run the multi-account setup script. Supports basic setup, custom config, custom permissions file, verification, and teardown. The teardown removes all created resources.

```bash
# Basic usage (uses default config: multi-cluster-config.yaml)
./setup-multi-account-iam.sh setup

# With custom config file
./setup-multi-account-iam.sh setup my-config.yaml

# With custom permissions file
./setup-multi-account-iam.sh setup my-config.yaml ./aws-mcp-iam-policy.json

# Verify the setup
./setup-multi-account-iam.sh verify my-config.yaml

# Teardown (removes all created resources)
./setup-multi-account-iam.sh teardown my-config.yaml

```

--------------------------------

### Install Azure API MCP server with go install

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/azure-mcp?q=

Install the Azure API MCP server using Go 1.24+ and rename the installed binary to azure-api-mcp for clarity. The binary is placed in $GOPATH/bin.

```bash
go install github.com/Azure/azure-api-mcp/cmd/server@latest
```

```bash
mv "$(go env GOPATH)/bin/server" "$(go env GOPATH)/bin/azure-api-mcp"
```

--------------------------------

### Quick start with Google Gemini

Source: https://holmesgpt.dev/latest/installation/cli-installation?q=

Quick start with Google Gemini. Set up the API key and model environment variables, then create a test pod to investigate. The TOOL_SCHEMA_NO_PARAM_OBJECT_IF_NO_PARAMS environment variable is required for Gemini.

```bash
export GEMINI_API_KEY="your-gemini-api-key"
export TOOL_SCHEMA_NO_PARAM_OBJECT_IF_NO_PARAMS=true
export MODEL="gemini/<your-gemini-model>"

```

```bash
kubectl apply -f https://raw.githubusercontent.com/robusta-dev/kubernetes-demos/main/pending_pods/pending_pod_node_selector.yaml

```

--------------------------------

### Apply and verify Azure MCP instance

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/azure-mcp?q=

Applies the manifest and verifies the deployment with kubectl get pods and logs. Uses the example name prod and namespace monitoring.

```bash
kubectl apply -f azure-mcp-additional-instance.yaml

# verify it comes up (use the name/namespace you chose in Step 2)
kubectl get pods -n monitoring -l app=azure-prod
kubectl logs  -n monitoring -l app=azure-prod

```

--------------------------------

### Quick start with OpenAI

Source: https://holmesgpt.dev/latest/installation/cli-installation?q=

Quick start with OpenAI. Set the API key, create a test pod, and ask your first question. The default model is gpt-4.1; you can specify a different model with --model.

```bash
export OPENAI_API_KEY="your-api-key"

```

```bash
kubectl apply -f https://raw.githubusercontent.com/robusta-dev/kubernetes-demos/main/pending_pods/pending_pod_node_selector.yaml

```

```bash
# Uses gpt-4.1 by default (fairly fast, decent results)
holmes ask "what is wrong with the user-profile-import pod?"

# Or specify a different model
holmes ask "what is wrong with the user-profile-import pod?" --model="gpt-5"

```

--------------------------------

### Install Dependencies

Source: https://holmesgpt.dev/latest/operator/development

Commands to install Python dependencies, activate the virtual environment, and install pre-commit hooks using Poetry.

```bash
# Install Python dependencies with Poetry
poetry install

# Activate virtual environment
poetry shell

# Install pre-commit hooks
poetry run pre-commit install
```

--------------------------------

### Install HolmesGPT from Source

Source: https://holmesgpt.dev/latest/installation/cli-installation

Clone the repository and install dependencies using Poetry for development builds.

```bash
git clone https://github.com/HolmesGPT/holmesgpt.git
cd holmesgpt
poetry install --no-root
```

```bash
poetry run holmes ask --help
```

--------------------------------

### Example response for /api/model

Source: https://holmesgpt.dev/latest/reference/http-api?q=

Example response for the /api/model GET request, showing a list of model names.

```json
{
  "model_name": ["anthropic/claude-sonnet-4-5-20250929", "anthropic/claude-opus-4-5-20251101", "robusta"]
}

```

--------------------------------

### Ollama quick start: create test pod, ask HolmesGPT with ollama_chat

Source: https://holmesgpt.dev/latest/installation/cli-installation?q=

Create a test pod and ask HolmesGPT to investigate it using the ollama_chat model. No API key is required for a local Ollama installation. Ollama can be tricky to configure; trying a hosted model first is recommended.

```bash
kubectl apply -f https://raw.githubusercontent.com/robusta-dev/kubernetes-demos/main/pending_pods/pending_pod_node_selector.yaml
```

```bash
holmes ask "what is wrong with the user-profile-import pod?" --model="ollama_chat/<your-model-name>"
```

--------------------------------

### Vertex AI quick start: set credentials, create test pod, ask HolmesGPT

Source: https://holmesgpt.dev/latest/installation/cli-installation?q=

Set up Google Vertex AI credentials, create a test pod, and ask HolmesGPT to investigate it using the vertex_ai model. The commands are part of the Vertex AI quick start workflow.

```bash
export VERTEXAI_PROJECT="your-project-id"
export VERTEXAI_LOCATION="us-central1"
export GOOGLE_APPLICATION_CREDENTIALS="path/to/service-account-key.json"
```

```bash
kubectl apply -f https://raw.githubusercontent.com/robusta-dev/kubernetes-demos/main/pending_pods/pending_pod_node_selector.yaml
```

```bash
holmes ask "what is wrong with the user-profile-import pod?" --model="vertex_ai/<your-vertex-model>"
```

--------------------------------

### Run setup script for GCP service account

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/gcp?q=

Recommended setup script that grants ~50 read-only roles and skips key generation. Provides audit logs, networking, security metadata, and monitoring access with read-only boundaries.

```bash
git clone https://github.com/robusta-dev/holmes-mcp-integrations.git
cd holmes-mcp-integrations/servers/gcp
./setup-gcp-service-account.sh --project ${PROJECT_ID} --skip-key-generation

```

--------------------------------

### Multi-account configuration YAML

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/aws?q=

Example configuration file for multi-account setup. Defines clusters, Kubernetes namespace/service account, IAM role/policy names, and target accounts with profiles. Edit this file with your cluster details and account IDs.

```yaml
clusters:
  - name: prod-cluster
    region: us-east-1
    account_id: "111111111111"
    oidc_issuer_id: AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
    oidc_issuer_url: https://oidc.eks.us-east-1.amazonaws.com/id/AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

  - name: staging-cluster
    region: us-west-2
    account_id: "111111111111"
    oidc_issuer_id: BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB
    oidc_issuer_url: https://oidc.eks.us-west-2.amazonaws.com/id/BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB

kubernetes:
  namespace: YOUR_NAMESPACE  # Must match the namespace where Holmes is deployed
  service_account: multi-account-mcp-sa

iam:
  role_name: EKSMultiAccountMCPRole
  policy_name: MCPReadOnlyPolicy
  session_duration: 3600

target_accounts:
  - profile: dev
    account_id: "111111111111"
    description: "Development account"

  - profile: prod
    account_id: "222222222222"
    description: "Production account"

```

--------------------------------

### Holmes OSS CLI: Configure skill_repos in config.yaml and run holmes ask

Source: https://holmesgpt.dev/latest/reference/skills?q=

For Holmes OSS CLI: add the skill repo to ~/.holmes/config.yaml with GitHub App credentials, then export the private key and run the holmes CLI. Holmes syncs the repo at the start of each run, minting an installation token from the private key.

```yaml
skill_repos:
  - url: https://github.com/<org>/<repo>.git
    branch: main        # optional; defaults to the repo's default branch
    sub_path: skills    # optional; subdirectory inside the repo where SKILL.md dirs live
    github_app_id: "<YOUR_APP_ID>"
    github_app_installation_id: "<YOUR_INSTALLATION_ID>"
    github_app_private_key_env: GITHUB_APP_PRIVATE_KEY

```

```bash
export GITHUB_APP_PRIVATE_KEY="$(cat /path/to/private-key.pem)"
holmes ask "why is my pod crashing?"

```

--------------------------------

### Install HolmesGPT via Homebrew

Source: https://holmesgpt.dev/latest/installation/cli-installation

Steps to add the tap, install the package, upgrade, and verify the installation.

```bash
brew tap robusta-dev/homebrew-holmesgpt
```

```bash
brew install holmesgpt
```

```bash
brew upgrade holmesgpt
```

```bash
holmes ask --help
```

--------------------------------

### Quick start with Azure AI Foundry

Source: https://holmesgpt.dev/latest/installation/cli-installation?q=

Quick start with Azure AI Foundry. Set the API key and endpoint environment variables, create a test pod, and ask your first question using the Azure model. See Azure AI Foundry Configuration for more details.

```bash
export AZURE_API_VERSION="2024-02-15-preview"
export AZURE_API_BASE="https://your-resource.openai.azure.com"
export AZURE_API_KEY="your-azure-api-key"

```

```bash
kubectl apply -f https://raw.githubusercontent.com/robusta-dev/kubernetes-demos/main/pending_pods/pending_pod_node_selector.yaml

```

```bash
holmes ask "what is wrong with the user-profile-import pod?" --model="azure/<your-model-name>"

```

--------------------------------

### Inline toolsets configuration example

Source: https://holmesgpt.dev/latest/development/evaluations/adding-evals

Inline toolsets configuration example showing how to enable kubernetes/core and disable aws/cloudwatch. This is the first method described under Advanced Features > Toolsets Configuration.

```yaml
toolsets:
  kubernetes/core:
    enabled: true
  aws/cloudwatch:
    enabled: false

```

--------------------------------

### Example usage: check missing indexes

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-postgresql

Prompt-style example asking for missing indexes on the orders table.

```text
"Check for missing indexes on the orders table"
```

--------------------------------

### Example prompt: show table schema

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-postgresql?q=

Example user prompt asking for the schema of a specific table. Useful for understanding table structure.

```text
"Show me the schema for the users table"

```

--------------------------------

### Install and verify vela CLI

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/kubevela

Install the KubeVela CLI using the official script and verify the installation with `vela version`.

```bash
# Install vela CLI
curl -fsSl https://kubevela.io/script/install.sh | bash

# Verify installation
vela version

```

--------------------------------

### Example Usage for Datadog Traces

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/datadog

Example commands to query Datadog traces using the Holmes CLI.

```bash
# Find slow requests
holmes ask "find traces where the checkout service took longer than 5 seconds"

# Analyze specific trace
holmes ask "analyze trace ID abc123 for performance issues"

# Service dependencies
holmes ask "show me traces involving both payment and inventory services"
```

--------------------------------

### Enable API server HTTPS (HOLMES_SSL_*)

Source: https://holmesgpt.dev/latest/reference/environment-variables?q=

Enable HTTPS for the HolmesGPT API. The CLI example sets the HOLMES_SSL_* environment variables; the Helm example enables TLS with a Kubernetes secret. Setting both HOLMES_SSL_CERTFILE and HOLMES_SSL_KEYFILE enables HTTPS; setting only one, or setting HOLMES_SSL_CA_CERTS or HOLMES_SSL_KEYFILE_PASSWORD without both server-side files, causes the server to fail to start. The certificate is read once at startup and does not hot-reload.

```bash
export HOLMES_SSL_CERTFILE=/path/to/tls.crt
export HOLMES_SSL_KEYFILE=/path/to/tls.key
# optional:
export HOLMES_SSL_KEYFILE_PASSWORD=changeit
export HOLMES_SSL_CA_CERTS=/path/to/ca.crt   # enables mTLS

```

```yaml
tls:
  enabled: true
  secretName: holmes-tls   # secret with tls.crt and tls.key
  # caCertsSecretKey: ca.crt   # optional: key in the secret, enables mTLS

```

--------------------------------

### Context Window Override Example

Source: https://holmesgpt.dev/latest/ai-providers/other

Example of using custom_args.max_context_size to manually set the context window to 1M tokens.

```yaml
modelList:
  my-model:
    model: provider/model-name
    api_key: "{{ env.API_KEY }}"
    custom_args:
      max_context_size: 1000000  # Override to 1M tokens
```

--------------------------------

### Example Usage for Datadog Metrics

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/datadog

Example commands to query Datadog metrics using the Holmes CLI.

```bash
# List available metrics
holmes ask "what metrics are available for my application?"

# Query CPU usage
holmes ask "show me CPU usage for the payment service over the last 6 hours"

# Custom application metrics
holmes ask "analyze the payment_processing_time metric for anomalies"
```

--------------------------------

### Example Datadog Traces Queries

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/datadog?q=

Example usage of the Datadog traces toolset. Shows how to query for slow requests, analyze a specific trace, and find service dependencies using the holmes ask command.

```bash
# Find slow requests
holmes ask "find traces where the checkout service took longer than 5 seconds"

# Analyze specific trace
holmes ask "analyze trace ID abc123 for performance issues"

# Service dependencies
holmes ask "show me traces involving both payment and inventory services"

```

--------------------------------

### Install HolmesGPT via pipx

Source: https://holmesgpt.dev/latest/installation/cli-installation

Install HolmesGPT using pipx and verify the installation.

```bash
pipx install holmesgpt
```

```bash
holmes ask --help
```

--------------------------------

### Command approval prompt example

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/bash

Example of the interactive prompt displayed when a command is not in the allow list. The user can choose to run once, approve the prefix, or provide feedback.

```text
Bash command

  kubectl scale deployment nginx --replicas=3
  Scale nginx deployment to 3 replicas

Do you want to proceed?
  1. Yes
  2. Yes, and don't ask again for `kubectl scale deployment nginx` commands
  3. Type here to tell Holmes what to do differently

```

--------------------------------

### Example MySQL usage: analyze slow query

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-mysql

Literal prompt example for the toolset: analyze a slow query on orders.

```text
"Analyze slow query: SELECT * FROM orders WHERE created_at > '2024-01-01'"


```

--------------------------------

### Install boto3 with minimum version

Source: https://holmesgpt.dev/latest/ai-providers/aws-bedrock?q=

Install boto3 version 1.28.57 or higher, required for AWS Bedrock integration.

```bash
pip install "boto3>=1.28.57"

```

--------------------------------

### List collections and schemas

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/mongodb

Example prompt for listing all collections in the orders database and displaying their schemas.

```text
"List all collections in the orders database and show their schemas"
```

--------------------------------

### Ask Questions Without Kubernetes

Source: https://holmesgpt.dev/latest/installation/cli-installation

Examples of querying HolmesGPT about infrastructure when not using a Kubernetes demo pod.

```bash
holmes ask "what Prometheus alerts are currently firing and why?"
holmes ask "what is the health of my Elasticsearch cluster?"
holmes ask "are there any issues with my production databases?"
```

--------------------------------

### Common Use Case: Suggest Indexes

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-sqlite?q=

Example query to suggest indexes for a table.

```sql
"Suggest indexes for the orders table"
```

--------------------------------

### Example response for /api/admin/reload/toolsets

Source: https://holmesgpt.dev/latest/reference/http-api?q=

Example response for the /api/admin/reload/toolsets POST request, showing status and counts of loaded toolsets and skills.

```json
{
  "status": "ok",
  "component": "toolsets",
  "detail": "50 toolsets loaded, 15 enabled, 12 skills",
  "counts": {
    "toolsets_total": 50,
    "toolsets_enabled": 15,
    "skills": 12
  }
}

```

--------------------------------

### Example response for /api/admin/reload

Source: https://holmesgpt.dev/latest/reference/http-api?q=

Example response for the /api/admin/reload POST request, showing status and counts of loaded components.

```json
{
  "status": "ok",
  "component": "all",
  "detail": "50 toolsets (15 enabled), 12 skills, 4 models",
  "counts": {
    "toolsets_total": 50,
    "toolsets_enabled": 15,
    "skills": 12,
    "models_loaded": 4
  }
}

```

--------------------------------

### Example NRQL Query for Application Logs

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/newrelic

An example of an automatically constructed NRQL query used by Holmes when investigating application logs.

```sql
SELECT message, timestamp FROM Log WHERE pod_name = 'your-app' SINCE 1 hour ago
```

--------------------------------

### Start interactive shell with /shell

Source: https://holmesgpt.dev/latest/reference/slash-commands

Starts an interactive shell session. When you exit, you can share the entire session with the AI and add a comment or question.

```text
> /shell
$ kubectl logs failing-pod-xyz
$ kubectl describe pod failing-pod-xyz
$ exit
Share shell session with AI? (Y/n): y
Add a comment or question (optional): The pod keeps failing with OOM errors

```

--------------------------------

### Start the Docker Compose server

Source: https://holmesgpt.dev/latest/installation/docker-compose-installation?q=

Starts the HolmesGPT HTTP server using Docker Compose. The server will be available at http://localhost:5050.

```bash
docker compose up

```

--------------------------------

### Example HolmesGPT database prompts

Source: https://holmesgpt.dev/latest/data-sources/database-connectors

Example natural-language prompts for HolmesGPT that use the database tools. Each prompt references a configured database instance (prod-postgres or analytics-clickhouse) and asks for a query, schema, table listing, or index check.

```text
"Why is this query slow on prod-postgres: SELECT * FROM orders WHERE status = 'pending'"

```

```text
"Show me the schema for the users table in analytics-clickhouse"

```

```text
"List all tables and their sizes in prod-postgres"

```

```text
"Check for missing indexes on the orders table"

```

--------------------------------

### PromQL Example Queries

Source: https://holmesgpt.dev/latest/reference/opentelemetry

Example PromQL queries for monitoring tool call rates and average LLM call duration in Prometheus or Grafana.

```promql
# Tool call rate by tool name
rate(holmesgpt_tool_call_count_total[5m])

# Average LLM call duration
rate(gen_ai_client_operation_duration_sum[5m]) / rate(gen_ai_client_operation_duration_count[5m])
```

--------------------------------

### MongoDB toolset example queries

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/mongodb?q=

Example natural-language queries for the MongoDB toolset. Each query maps to one of the available tools (e.g., list_collections, find, aggregate, current_op, server_status). Use these as prompts to exercise the corresponding tool.

```text
"List all collections in the orders database and show their schemas"
```

```text
"Find the 10 most recent error documents in the events collection"
```

```text
"How many orders per status are there in the last 24 hours?"
```

```text
"Are there any slow queries running on the MongoDB server?"
```

```text
"Check the MongoDB server health — connections, memory, and lock statistics"
```

--------------------------------

### Dynatrace DQL Example Queries

Source: https://holmesgpt.dev/latest/reference/opentelemetry

Example DQL queries for analyzing tool call duration, token usage, investigation counts, and tool performance in Dynatrace.

```sql
# Tool call duration by tool name
timeseries avg(holmesgpt.tool.call.duration, default:0), by: {holmesgpt_tool_name}

# Token usage by model
timeseries sum(gen_ai.client.token.usage, default:0), by: {gen_ai_request_model, gen_ai_token_type}

# Investigation count over time
timeseries sum(holmesgpt.investigation.count, default:0)

# Slowest tools (p95)
timeseries percentile(holmesgpt.tool.call.duration, 95), by: {holmesgpt_tool_name}
```

--------------------------------

### SQLite query: suggest indexes

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-sqlite

Example natural-language query to get index suggestions for a table.

```text
"Suggest indexes for the orders table"

```

--------------------------------

### Example MySQL usage: find largest tables

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-mysql

Literal prompt to get 10 largest tables.

```text
"What are the 10 largest tables?"


```

--------------------------------

### Prompt: Show table structure and suggest indexes

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-mysql?q=

Example prompt to show table structure for products and suggest indexes.

```text
"Show table structure for products and suggest indexes"

```

--------------------------------

### Piped command example: kubectl get pods | grep error | head -10

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/bash?q=

Each segment of a piped command is validated independently. For this command to run, `kubectl get`, `grep`, and `head` must all be in the allow list.

```bash
kubectl get pods | grep error | head -10

```

--------------------------------

### holmes ask examples for Kubernetes remediation

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/kubernetes-remediation-mcp?q=

Four example invocations of the `holmes ask` CLI for common Kubernetes remediation tasks: reading a file from a pod, checking DNS/endpoint reachability, restarting a deployment, and investigating/fixing crashlooping pods. Each command is a single-line shell invocation.

```bash
holmes ask "Read /app/config.yaml from the checkout-api pod and tell me what database host it points to"

```

```bash
holmes ask "From inside the production cluster, check whether the payments service DNS resolves and the endpoint is reachable"

```

```bash
holmes ask "Restart the payment-service deployment in the production namespace"

```

```bash
holmes ask "The checkout-api pods are crashlooping - investigate and fix"

```

--------------------------------

### Example Usage of Datadog Toolset

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/datadog

Use the holmes ask command to query various Datadog resources.

```bash
# List all monitors
holmes ask "show me all Datadog monitors"

# Get dashboard details
holmes ask "retrieve my application dashboard from Datadog"

# Check SLO status
holmes ask "what's the current status of our API availability SLO?"

# Search incidents
holmes ask "find recent incidents in Datadog"

# Get synthetic test results
holmes ask "show me the latest synthetic test results for our homepage"
```

--------------------------------

### Endpoint whitelist hosts patterns example

Source: https://holmesgpt.dev/latest/data-sources/api-toolsets?q=

Example of the hosts list in the endpoint whitelist. Shows bare hostnames, wildcards, scheme+port pins, and the :* wildcard for any port. Malformed entries fail at config-load time.

```yaml
hosts:
  - "api.example.com"                    # any scheme, any port
  - "*.internal.example.com"             # any subdomain, any scheme, any port
  - "https://jenkins.example.com:8080"   # https on 8080 only
  - "https://*.tools.example.com:*"      # https on any port, any subdomain

```

--------------------------------

### Analyze Execution Plan Use Case

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-sqlserver?q=

Example prompt for analyzing a query execution plan.

```text
"Analyze execution plan for: SELECT * FROM Orders WHERE CustomerId = 123"
```

--------------------------------

### Configure and Use AWS Bedrock

Source: https://holmesgpt.dev/latest/installation/cli-installation

Set AWS credentials, install boto3, create a test pod, and run a query.

```bash
export AWS_ACCESS_KEY_ID="your-access-key"
export AWS_SECRET_ACCESS_KEY="your-secret-key"
export AWS_DEFAULT_REGION="your-region"
```

```bash
pip install "boto3>=1.28.57"
```

```bash
kubectl apply -f https://raw.githubusercontent.com/robusta-dev/kubernetes-demos/main/pending_pods/pending_pod_node_selector.yaml
```

```bash
holmes ask "what is wrong with the user-profile-import pod?" --model="bedrock/us.anthropic.claude-sonnet-4-5-20250929-v1:0"

# Or use another model
holmes ask "what is wrong with the user-profile-import pod?" --model="bedrock/<your-model-name>"
```

--------------------------------

### Custom toolsets.yaml with prometheus/metrics

Source: https://holmesgpt.dev/latest/development/evaluations/adding-evals

Example toolsets.yaml file that enables prometheus/metrics with a custom prometheus_url and disables grafana/dashboards. This is the preferred way to configure toolsets for complex setups.

```yaml
toolsets:
  prometheus/metrics:
    enabled: true
    config:
      prometheus_url: "http://custom-prometheus:9090"
  grafana/dashboards:
    enabled: false  # Disable specific toolsets

```

--------------------------------

### Image Analysis with URL

Source: https://holmesgpt.dev/latest/reference/http-api

Example request providing an image via a direct URL.

```bash
curl -X POST http://<HOLMES-URL>/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "ask": "What is in this image?",
    "model": "anthropic/claude-sonnet-4-5-20250929",
    "images": [
      "https://example.com/screenshot.png"
    ]
  }'
```

--------------------------------

### Quick start with Anthropic Claude

Source: https://holmesgpt.dev/latest/installation/cli-installation?q=

Quick start with Anthropic Claude. Set the API key, create a test pod, and ask your first question using the specified model. You can change the model name to any Anthropic model.

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

```

```bash
kubectl apply -f https://raw.githubusercontent.com/robusta-dev/kubernetes-demos/main/pending_pods/pending_pod_node_selector.yaml

```

```bash
holmes ask "what is wrong with the user-profile-import pod?" --model="anthropic/claude-sonnet-4-5-20250929"

```

--------------------------------

### Example Holmes CLI queries for Datadog general toolset

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/datadog?q=

Example usage of the Datadog general toolset via the Holmes CLI. These commands ask Holmes to list monitors, retrieve dashboard details, check SLO status, search incidents, and get synthetic test results.

```bash
# List all monitors
holmes ask "show me all Datadog monitors"

# Get dashboard details
holmes ask "retrieve my application dashboard from Datadog"

# Check SLO status
holmes ask "what's the current status of our API availability SLO?"

# Search incidents
holmes ask "find recent incidents in Datadog"

# Get synthetic test results
holmes ask "show me the latest synthetic test results for our homepage"

```

--------------------------------

### Run setup-azure-identity.sh for Azure RBAC

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/azure-mcp?q=

Downloads and runs the recommended setup script for Azure RBAC. It creates a managed identity, assigns RBAC roles, configures federated credentials, and outputs the configuration values for the Helm chart. Requires the script to be executed with the specified flags.

```bash
curl -O https://raw.githubusercontent.com/robusta-dev/holmes-mcp-integrations/master/servers/azure/setup-azure-identity.sh
bash setup-azure-identity.sh --auth-method workload-identity \
  --resource-group YOUR_RESOURCE_GROUP \
  --aks-cluster YOUR_AKS_CLUSTER \
  --all-subscriptions

```

--------------------------------

### Holmes OSS Helm Chart: GCP service account setup and deployment

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/gcp?q=

For Holmes OSS Helm Chart: clone the integration repo, run the setup script with your project and namespace, add the gcp MCP addon configuration to values.yaml, then deploy with helm upgrade.

```bash
git clone https://github.com/robusta-dev/holmes-mcp-integrations.git
cd holmes-mcp-integrations/servers/gcp

./setup-gcp-service-account.sh \
  --project your-project-id \
  --k8s-namespace holmes

```

```yaml
mcpAddons:
  gcp:
    enabled: true
    serviceAccountKey:
      secretName: "gcp-sa-key"
    # Optional: defaults when user doesn't specify. Holmes can query any project the SA has access to.
    config:
      project: "your-primary-project"
      region: "us-central1"
    gcloud:
      enabled: true
    observability:
      enabled: true
    storage:
      enabled: true

```

```bash
helm upgrade --install holmes robusta/holmes -f values.yaml

```

--------------------------------

### Configure Host Patterns

Source: https://holmesgpt.dev/latest/data-sources/api-toolsets

Example configuration for the hosts list, demonstrating various combinations of schemes, subdomains, and ports.

```yaml
hosts:
  - "api.example.com"                    # any scheme, any port
  - "*.internal.example.com"             # any subdomain, any scheme, any port
  - "https://jenkins.example.com:8080"   # https on 8080 only
  - "https://*.tools.example.com:*"      # https on any port, any subdomain
```

--------------------------------

### Show Database Size Use Case

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-sqlserver?q=

Example prompt for retrieving database size and file growth settings.

```text
"Show database size and file growth settings"
```

--------------------------------

### GET /api/model - List available models

Source: https://holmesgpt.dev/latest/reference/http-api

Returns a list of available AI models that can be used for chat. Example shows curl command and JSON response with model names.

```bash
curl http://<HOLMES-URL>/api/model

```

```json
{
  "model_name": ["anthropic/claude-sonnet-4-5-20250929", "anthropic/claude-opus-4-5-20251101", "robusta"]
}

```

--------------------------------

### Quick Start: Configure and Ask a Question

Source: https://holmesgpt.dev/latest/reference/python-sdk

Sets up the HolmesGPT client with an Anthropic API key, creates an AI instance with core and CLI toolsets, and asks a question about failing pods. Requires the ANTHROPIC_API_KEY environment variable.

```python
import os
from holmes.config import Config
from holmes.core.prompt import build_initial_ask_messages
from holmes.core.tools import ToolsetTag

# Create configuration
config = Config(
    api_key=os.getenv("ANTHROPIC_API_KEY"),
    model="anthropic/claude-sonnet-4-5-20250929",
)

# Create AI instance
ai = config.create_toolcalling_llm(
    # Only load toolsets tagged CORE or CLI (excludes server-only CLUSTER toolsets)
    toolset_tag_filter=[ToolsetTag.CORE, ToolsetTag.CLI],
    # Auto-enable every toolset that works without explicit config (e.g. kubectl on PATH)
    enable_all_toolsets_possible=True,
    # Remaining params use defaults:
    #   prerequisite_cache=PrerequisiteCacheMode.ENABLED — use cached health-check results
    #   reuse_executor=False       — create a fresh executor each call (fine for CLI)
)

# Ask a question
messages = build_initial_ask_messages(
    initial_user_prompt="what pods are failing in production?",
    file_paths=None,
    tool_executor=ai.tool_executor,
    runbooks=config.get_runbook_catalog(),
    system_prompt_additions=None,
)

response = ai.call(messages)
print(response.result)
```

--------------------------------

### Get Confluence page at URL and summarize remediation steps

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/confluence?q=

Use this example to fetch a specific Confluence page by URL and summarize its remediation steps. The confluence_request capability supports fetching pages.

```text
Get the Confluence page at https://mycompany.atlassian.net/wiki/spaces/OPS/pages/12345 and summarize the remediation steps
```

--------------------------------

### Minimal prompt behavior_controls example

Source: https://holmesgpt.dev/latest/reference/http-api

Demonstrates how to disable most prompt sections to reduce token usage and latency.

```bash
curl -X POST http://<HOLMES-URL>/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "ask": "List all pods in the default namespace",
    "behavior_controls": {
      "todowrite_instructions": false,
      "todowrite_reminder": false,
      "ai_safety": false,
      "style_guide": false,
      "general_instructions": false
    }
  }'
```

--------------------------------

### Launch Interactive Mode

Source: https://holmesgpt.dev/latest/walkthrough/interactive-mode

Start the interactive session by running the ask command without arguments.

```bash
holmes ask
```

--------------------------------

### Listing Available Tools

Source: https://holmesgpt.dev/latest/reference/python-sdk

Creates an AI instance with core and CLI toolsets, then prints the loaded toolsets with their enabled status and the names of individual tools.

```python
ai = config.create_toolcalling_llm(
    toolset_tag_filter=[ToolsetTag.CORE, ToolsetTag.CLI],
    enable_all_toolsets_possible=True,
)

# List loaded toolsets and their status
for toolset in ai.tool_executor.toolsets:
    print(f"  {toolset.name} ({'enabled' if toolset.enabled else 'disabled'})")

# List individual tools
for tool_name in sorted(ai.tool_executor.tools_by_name.keys()):
    print(f"  {tool_name}")
```

--------------------------------

### Find on-call runbook in SRE space and get escalation contacts

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/confluence?q=

Use this example to find the on-call runbook in the SRE space and retrieve escalation contacts. The confluence_request capability can list spaces and fetch pages.

```text
Find the on-call runbook in the SRE space and tell me the escalation contacts
```

--------------------------------

### Correct API Model Usage

Source: https://holmesgpt.dev/latest/reference/http-api

Example of using the modelList key name in an API request.

```bash
# Use the modelList key name, not the actual model identifier
curl -X POST http://localhost:8080/api/chat \
  -H "Content-Type: application/json" \
  -d '{"ask": "list pods", "model": "fast-model"}'
```

--------------------------------

### Define my_toolset.yaml

Source: https://holmesgpt.dev/latest/reference/python-sdk?q=

Example YAML structure for defining custom tools, including descriptions, prerequisites, and shell commands.

```yaml
toolsets:
  my-service-tools:
    description: "Tools for checking my custom service"
    prerequisites:
      - command: "curl --version"
    tools:
      - name: check_service_health
        description: "Check the health endpoint of my service"
        command: |
          curl -s "${MY_SERVICE_URL}/health"

      - name: get_service_metrics
        description: "Get Prometheus-style metrics from my service"
        command: |
          curl -s "${MY_SERVICE_URL}/metrics" | head -50
```

--------------------------------

### Configure and Use OpenAI

Source: https://holmesgpt.dev/latest/installation/cli-installation

Set the API key, create a test pod, and run a query using default or custom OpenAI models.

```bash
export OPENAI_API_KEY="your-api-key"
```

```bash
kubectl apply -f https://raw.githubusercontent.com/robusta-dev/kubernetes-demos/main/pending_pods/pending_pod_node_selector.yaml
```

```bash
# Uses gpt-4.1 by default (fairly fast, decent results)
holmes ask "what is wrong with the user-profile-import pod?"

# Or specify a different model
holmes ask "what is wrong with the user-profile-import pod?" --model="gpt-5"
```

--------------------------------

### Customize Sentry MCP LLM instructions in Robusta Helm Chart

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/sentry-mcp?q=

Customize how Holmes uses Sentry by providing your own LLM instructions in the mcpAddons.sentry.llmInstructions field. The example instructs Holmes to start by listing projects, search for relevant issues, and retrieve full stack traces before drawing conclusions.

```yaml
mcpAddons:
  sentry:
    enabled: true
    auth:
      secretName: "sentry-mcp-token"
    llmInstructions: |
      Use the Sentry MCP to investigate application errors and crashes.
      When investigating, always start by listing projects, then search for relevant issues,
      and retrieve full stack traces before drawing conclusions.

```

--------------------------------

### Build Azure API MCP from source

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/azure-mcp?q=

Build the Azure API MCP server from source by cloning the repository, building with Go, and installing the binary to /usr/local/bin.

```bash
git clone https://github.com/Azure/azure-api-mcp.git
cd azure-api-mcp
go build -o azure-api-mcp ./cmd/server
sudo mv azure-api-mcp /usr/local/bin/
```

--------------------------------

### Custom Python toolset for httpbin IP lookup

Source: https://holmesgpt.dev/latest/reference/python-sdk

Complete example of a custom Python toolset that calls httpbin.org to get the caller's IP address. It defines a Pydantic config class, a Tool subclass implementing _invoke() and get_parameterized_one_liner(), and a Toolset subclass with a prerequisites_callable() health check. The example then passes the toolset to the SDK via additional_toolsets and runs a query. Note that _toolset must be set after super().__init__() in tool classes, and _invoke() returns StructuredToolResult with SUCCESS or ERROR status.

```python
import os
import requests
from typing import Any, ClassVar, Dict, List, Tuple, Type
from pydantic import BaseModel
from holmes.config import Config
from holmes.core.prompt import build_initial_ask_messages
from holmes.core.tools import (
    CallablePrerequisite,
    StructuredToolResult,
    StructuredToolResultStatus,
    Tool,
    ToolInvokeContext,
    Toolset,
    ToolsetTag,
)


# 1. Config class
class HttpBinConfig(BaseModel):
    api_url: str = "https://httpbin.org"


# 2. Tool class - implements the actual HTTP call
class GetIp(Tool):
    def __init__(self, toolset: "HttpBinToolset"):
        super().__init__(
            name="httpbin_get_ip",
            description="Get the caller's public IP address via httpbin.org",
            parameters={},
        )
        # Important: set _toolset AFTER super().__init__() (Pydantic requirement)
        self._toolset = toolset

    def _invoke(self, params: dict, context: ToolInvokeContext) -> StructuredToolResult:
        resp = requests.get(f"{self._toolset.get_config().api_url}/ip", timeout=10)
        if resp.status_code != 200:
            return StructuredToolResult(
                status=StructuredToolResultStatus.ERROR,
                data=f"HTTP {resp.status_code}: {resp.text}",
                params=params,
            )
        return StructuredToolResult(
            status=StructuredToolResultStatus.SUCCESS,
            data=resp.json(),
            params=params,
        )

    def get_parameterized_one_liner(self, params: Dict) -> str:
        return "httpbin: Get IP address"


# 3. Toolset class - groups tools + health check
class HttpBinToolset(Toolset):
    config_classes: ClassVar[List[Type[HttpBinConfig]]] = [HttpBinConfig]

    def __init__(self):
        super().__init__(
            name="httpbin",
            description="Tools for testing HTTP requests via httpbin.org",
            enabled=True,
            prerequisites=[CallablePrerequisite(callable=self.prerequisites_callable)],
            tools=[GetIp(self)],
            tags=[ToolsetTag.CORE],
        )
        self._httpbin_config = HttpBinConfig()

    def get_config(self) -> HttpBinConfig:
        return self._httpbin_config

    def prerequisites_callable(self, config: dict[str, Any]) -> Tuple[bool, str]:
        try:
            self._httpbin_config = HttpBinConfig(**config) if config else HttpBinConfig()
            resp = requests.get(f"{self._httpbin_config.api_url}/get", timeout=5)
            if resp.ok:
                return True, "httpbin.org reachable"
            return False, f"Health check failed: HTTP {resp.status_code}"
        except Exception as e:
            return False, f"Cannot reach httpbin.org: {e}"


# 4. Use it - pass the toolset via additional_toolsets
config = Config(
    api_key=os.getenv("ANTHROPIC_API_KEY"),
    model="anthropic/claude-sonnet-4-5-20250929",
    additional_toolsets=[HttpBinToolset()],
)

ai = config.create_toolcalling_llm(
    toolset_tag_filter=[ToolsetTag.CORE, ToolsetTag.CLI],
    enable_all_toolsets_possible=True,
)

messages = build_initial_ask_messages(
    initial_user_prompt="What is my public IP address?",
    file_paths=None,
    tool_executor=ai.tool_executor,
    runbooks=config.get_runbook_catalog(),
    system_prompt_additions=None,
)
response = ai.call(messages)
print(response.result)
```

--------------------------------

### Configure Azure SQL Connection

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-sqlserver?q=

Example configuration for the database toolset using the mssql+pytds driver.

```yaml
toolsets:
  azure-sql-prod:
    type: database
    config:
      connection_url: "mssql+pytds://holmes_readonly:Your_Secure_Password123!@yourserver.database.windows.net:1433/mydb"
    llm_instructions: "Production Azure SQL database with application data"
```

--------------------------------

### Launch Interactive Mode with Initial Question

Source: https://holmesgpt.dev/latest/walkthrough/interactive-mode

Start the interactive session and immediately provide an initial query for the AI to investigate.

```bash
holmes ask "what pods are failing?"
```

--------------------------------

### Query Azure cost spikes

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/azure-mcp

Example prompt for analyzing sudden increases in Azure cloud spending.

```text
"Our Azure costs increased 50% last week"
```

--------------------------------

### Create GCP Service Account (Robusta Helm Chart)

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/gcp

Clones the holmes-mcp-integrations repository and runs the setup script to create a service account with ~50 read-only IAM roles, generate a key, and create a Kubernetes secret named gcp-sa-key in the robusta namespace.

```bash
git clone https://github.com/robusta-dev/holmes-mcp-integrations.git
cd holmes-mcp-integrations/servers/gcp

./setup-gcp-service-account.sh \
  --project your-project-id \
  --k8s-namespace robusta

```

--------------------------------

### Install HolmesGPT dependencies

Source: https://holmesgpt.dev/latest/development/evaluations/running-evals

Install the necessary Python dependencies for HolmesGPT using poetry.

```bash
poetry install --with=dev
```

--------------------------------

### Quick Start with HolmesGPT Python SDK

Source: https://holmesgpt.dev/latest/installation/python-installation

Configures the HolmesGPT SDK with an Anthropic API key, creates a tool-calling LLM with core and CLI toolsets, builds initial ask messages for a user prompt, and prints the response.

```python
import os
from holmes.config import Config
from holmes.core.prompt import build_initial_ask_messages
from holmes.core.tools import ToolsetTag

config = Config(
    api_key=os.getenv("ANTHROPIC_API_KEY"),
    model="anthropic/claude-sonnet-4-5-20250929",
)

ai = config.create_toolcalling_llm(
    toolset_tag_filter=[ToolsetTag.CORE, ToolsetTag.CLI],
    enable_all_toolsets_possible=True,
)

messages = build_initial_ask_messages(
    initial_user_prompt="what pods are failing in production?",
    file_paths=None,
    tool_executor=ai.tool_executor,
    runbooks=config.get_runbook_catalog(),
    system_prompt_additions=None,
)

response = ai.call(messages)
print(response.result)
```

--------------------------------

### Configure and Use Google Gemini

Source: https://holmesgpt.dev/latest/installation/cli-installation

Set Gemini environment variables and create a test pod.

```bash
export GEMINI_API_KEY="your-gemini-api-key"
export TOOL_SCHEMA_NO_PARAM_OBJECT_IF_NO_PARAMS=true
export MODEL="gemini/<your-gemini-model>"
```

```bash
kubectl apply -f https://raw.githubusercontent.com/robusta-dev/kubernetes-demos/main/pending_pods/pending_pod_node_selector.yaml
```

--------------------------------

### Install HolmesGPT

Source: https://holmesgpt.dev/latest/installation/kubernetes-installation?q=

Command to install the HolmesGPT Helm chart using the configured values.yaml file.

```bash
helm install holmesgpt robusta/holmes -f values.yaml
```

--------------------------------

### Apply the manifest and verify deployment

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/azure-mcp

Apply the configured manifest to the cluster and verify that the new pod is running correctly.

```bash
kubectl apply -f azure-mcp-additional-instance.yaml

# verify it comes up (use the name/namespace you chose in Step 2)
kubectl get pods -n monitoring -l app=azure-prod
kubectl logs  -n monitoring -l app=azure-prod
```

--------------------------------

### Install HolmesGPT dev dependencies

Source: https://holmesgpt.dev/latest/development/evaluations/adding-evals?q=

Install the Python development dependencies required for running evals.

```bash
poetry install --with=dev

```

--------------------------------

### Common Use Case: Show Tables and Row Counts

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-sqlite?q=

Example query to show all tables and their row counts.

```sql
"Show all tables and their row counts"
```

--------------------------------

### Test ArgoCD Toolset

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/argocd

Sample question to verify the ArgoCD toolset works.

```bash
holmes ask "Which ArgoCD applications are failing and why?"
```

--------------------------------

### Example usage: analyze slow query

Source: https://holmesgpt.dev/latest/data-sources/builtin-toolsets/database-postgresql

Example free-text prompt for HolmesGPT to investigate a slow query.

```text
"Why is this query slow: SELECT * FROM orders WHERE customer_id = 12345"
```

--------------------------------

### Install or Upgrade Helm Release

Source: https://holmesgpt.dev/latest/operator/development

Commands to install, upgrade, and verify the operator deployment using the local configuration.

```bash
# Install new release
helm install holmesgpt helm/holmes -f values-dev.yaml

# Upgrade existing release
helm upgrade holmesgpt helm/holmes -f values-dev.yaml

# Verify deployment
kubectl get pods -l app.kubernetes.io/name=holmes-operator
kubectl logs -l app.kubernetes.io/name=holmes-operator --tail=50
```

--------------------------------

### Install Unreleased Version from GitHub

Source: https://holmesgpt.dev/latest/installation/python-installation

Installs an unreleased version of the HolmesGPT package directly from the master branch on GitHub.

```bash
pip install "https://github.com/HolmesGPT/holmesgpt/archive/refs/heads/master.zip"
```

--------------------------------

### Install HolmesGPT Python Package

Source: https://holmesgpt.dev/latest/installation/python-installation

Installs the latest stable version of the HolmesGPT Python package using pip.

```bash
pip install holmesgpt # Installs latest stable version
```