Explore documentation
AI SRE vs MCP server
Better Stack offers two ways to use AI on your telemetry data: the browser-based AI SRE and the IDE-focused MCP server. Both can be used at the same time.
You can configure the MCP server from AI SRE → MCP and API.
| Aspect | AI SRE | MCP server |
|---|---|---|
| Where you use it | In your browser | In your IDE or a compatible agent |
| Who runs the model | Better Stack | Your own LLM provider |
| Who pays for tokens | Better Stack bills your organization | You pay your LLM provider directly |
| What it can do | Full toolset: code, dashboards, alerts | Same data access as AI SRE; without the code-writing, dashboard-editing, or alert-managing tools. |
| Authentication | Your Better Stack login | OAuth or Better Stack API token |
| Setup | None, just open the chat | Configured in your MCP client |
Advantages of AI SRE
- Zero setup: Open the chat and start asking questions. Nothing to install.
- Most powerful tool set: Only AI SRE can clone your repos, write code, open pull requests, and manage dashboards and alerts.
- Integrations baked in: Once connected, AI SRE can automatically use tools like Sentry, Datadog, and Linear during an investigation.
- Built-in awareness: AI SRE always knows the current time, team, on-call schedule, and ongoing incidents.
- Conversation history: Every investigation is saved and shareable within your team.
Advantages of MCP server
- Use your own model: If you have an existing account with Anthropic, OpenAI, or another provider, you can use your preferred model.
- Use your own tools: Combine Better Stack data with anything else your AI agent can access, such as your local codebase.
- Automate debugging: Build custom workflows like bulk resolving errors without leaving your IDE.
- No Better Stack billing: Token costs are billed directly by your LLM provider.
Who pays for tokens
| Aspect | AI SRE | MCP server |
|---|---|---|
| Token cost | Charged to your Better Stack organization. Visible in Settings → Usage. | Charged to your LLM provider account, e.g. Anthropic, OpenAI. |
| Spending limit | If your organization hits its Better Stack spending limit, AI SRE is automatically gated until the next billing cycle. | Independent of Better Stack billing. |
| Plan requirement | Requires a plan that includes AI SRE as a paid product. Free plans see an upsell prompt. | Requires a plan that includes MCP server access. |
Privileges
AI SRE inherits your role and permissions on the team. If you are a read-only member, AI SRE cannot perform actions that modify state (like creating alerts or opening pull requests) on your behalf and will inform you that it lacks permission. For example, creating alerts requires a Responder role or higher. For a full list of permissions, see the team roles documentation.
For sensitive actions such as writing code or creating alerts, AI SRE will show a Pending confirmation card in the chat. You can approve the action, deny it, or approve it for the rest of the session.