- Newest
- Most votes
- Most comments
Prerequisites
AWS Side
- An active AWS DevOps Agent Agent Space
- Access to the AWS DevOps Agent console (Capability Providers section)
Observe Side
- An Observe account with AI features enabled
- The Observe MCP endpoint URL: https://<customer-id>.observeinc.com/v1/ai/mcp
- A valid API token (generated via Observe UI: Manage Account → My API Tokens)
Note: If the user is unable to create an API token, their Observe group needs the "Authtokens → Create" permission. This can be granted via a custom group with just that checkbox selected, or by assigning the built-in writer role.
Observe Knowledge Graph
- Observe maintains a Knowledge Graph (AI-first vector store) that indexes datasets, metrics, traces, and their relationships
- This is auto-populated — no manual intervention needed
- MCP tools query this Knowledge Graph, not the raw datasets directly
- Once AI features are enabled and data is flowing, the Knowledge Graph will be populated automatically
Step-by-Step Integration
Step 1: Obtain the Observe MCP Endpoint URL
Your Observe MCP endpoint follows this format:
https://<customer-id>.observeinc.com/v1/ai/mcp
To find your customer ID:
- Log into Observe
- Navigate to Admin → Customer Settings
- The "ID" field is your customer ID (e.g., 123456789012)
Alternatively, check your browser URL — it will be https://<customer-id>.observeinc.com/...
Step 2: Generate an API Token
- In Observe, click your username in the left navigation rail → Manage account
- Click "My API tokens"
- Click "Create API Token"
- Name: devops-agent-mcp
- Set expiration as appropriate
- Copy the token value
Your authorization header value will be: Bearer <customer-id> <token>
Step 3: Verify the MCP Endpoint is Reachable
Test connectivity from your terminal:
curl -v -X POST https://<customer-id>.observeinc.com/v1/ai/mcp
-H "Content-Type: application/json"
-H "Authorization: Bearer <customer-id> <token>"
-d '{"jsonrpc":"2.0","method":"initialize","id":1,"params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}'
Expected response: A valid JSON-RPC response (HTTP 200 or 406). Any response from the server confirms the endpoint is active.
Step 4: Register the MCP Server in DevOps Agent
Via Console:
- Open the AWS DevOps Agent console
- Navigate to Capability Providers
- Find "MCP Server" → Click Register
- Fill in:
- Server Name: observe-mcp
- Endpoint URL: https://<customer-id>.observeinc.com/v1/ai/mcp
- Check "Enable Dynamic Client Registration"
- Auth type: API Key
- API Key Header: Authorization
- API Key Value: Bearer <customer-id> <token>
- Click Register
Via AWS CLI:
aws devops-agent register-service
--service mcpserver
--service-details '{
"mcpserver": {
"name": "observe-mcp",
"endpoint": "https://<endpoint>/v1/ai/mcp",
"description": "Observe observability platform MCP server",
"authorizationConfig": {
"apiKey": {
"apiKeyName": "observe-api-key",
"apiKeyValue": "Bearer <customer-id> <token>",
"apiKeyHeader": "Authorization"
}
}
}
}'
--region us-east-1
Step 5: Select MCP Tools
After registration, DevOps Agent will discover available tools. For Observe, these include:
| Tool | Purpose |
|---|---|
| observe-skill | General Observe query capability |
| observe-dataset-search | Find/discover datasets |
| observe-metric-search | Search metrics |
| observe-correlation-tag-value-search | Search tag values |
| observe-correlation-tag-key-search | Search tag keys |
| observe-generate-query-card | Generate log/metric queries |
| observe-generate-service-map | Visualize service dependencies |
| observe-search-monitor-alerts | Search triggered alerts |
| observe-list-monitors | List configured monitors |
| observe-get-monitor | Get details of a specific monitor |
All 10 tools are read-only and safe to enable. Select all and save.
Step 6: Add MCP to Your Agent Space
- Go to your Agent Space → View Details
- Select the Capabilities tab
- Under MCP Servers or Telemetry, click Add
- Select observe-mcp → Save
- Note the Webhook URL and Secret (optional — for Observe to auto-trigger investigations)
Step 7: Test the Integration
Open your Agent Space chat and ask:
- "List monitors in Observe"
- "Search for metrics in Observe"
- "Generate a service map from Observe"
- "Search for recent alerts in Observe"
To integrate Observe with AWS DevOps Agent using a custom MCP server, you'll need to follow the bring-your-own MCP server approach. This allows you to connect observability tools beyond the built-in integrations through a standardized protocol that AWS DevOps Agent can query during investigations.
Here are the key requirements for setting up a custom MCP server for Observe:
Technical Requirements:
- Your MCP server must have a publicly accessible HTTPS endpoint that is reachable from the public internet. VPC-hosted servers are not currently supported.
- The MCP server should only expose read-only tools for security purposes. Write operations introduce prompt injection risks and should not be included.
- You'll need to register the MCP server at the account level first, then selectively enable specific tools per Agent Space. It's important not to grant access to all tools—only choose those relevant to your investigations.
Registration Process: Once you have your Observe MCP server endpoint ready, you would register it as a service using the AWS DevOps Agent CLI with a command structure similar to other MCP server integrations. You'll need to provide the service name, endpoint URL, and appropriate authentication configuration (such as bearer token credentials).
After registration, you'll receive a service ID that you can then use to associate the MCP server with your specific Agent Space. This association step includes configuring the endpoint details and enables AWS DevOps Agent to query your Observe data during investigations.
The agent will then be able to access your Observe telemetry data on-demand during incident investigations, correlating it with AWS service data and other connected observability tools without requiring custom integration pipelines or data aggregation layers.
Sources
Best Practices for Deploying AWS DevOps Agent in Production | AWS DevOps & Developer Productivity Blog
AWS DevOps Agent CLI onboarding guide - AWS DevOps Agent
Relevant content
asked 4 months ago
- AWS OFFICIALUpdated 4 years ago
