HolmesGPT

https://holmesgpt.dev/latest/
Admin
HolmesGPT is an open-source AI agent for investigating production incidents and automating...
Tokens:685,269
Snippets:4,336
Trust Score:6.6
Update:1 month ago
Show doc for...
Tokens:
Raw
### 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"
```

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

### 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"
```

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

### 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"]
}

```

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

### 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

```

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

### 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"
```

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

### 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>"
```

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

### 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"

```

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

### 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"

```

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

### 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

```

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

### 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

```

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

### 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>"

```

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

### 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"

```

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

### 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

```

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

### 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

```

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

### 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
```

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

### 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

```

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

### 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"]
}

```

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

### 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
```

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

### 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)
```

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

### 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
```

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

### 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

```
Tokens:
Raw