Last9 AI

Last9 AI investigates production issues using your logs, metrics, traces, and alerts. Ask about a service and a time window, and it queries your telemetry, reports the evidence behind each finding, and links back to the underlying data in Last9.

It runs on Mithai, Last9’s open-source multiplayer agent harness for your organization. Mithai supplies the sessions, memory, skills, and chat adapters behind every surface, so a Slack thread holds one investigation that several engineers work in together. You can also run Mithai yourself for workflows that Last9 AI does not cover.

What Last9 AI does

• Investigates across signals. It follows an error from an exception to the failing trace and the correlated logs, and checks change events near the incident.
• Shows its evidence. Answers include links to the queries behind them, and they separate confirmed findings from possible causes.
• Works with your team. Several engineers investigate together in one Slack thread. An AI Assistant conversation is your own, and you can share it with a colleague to read. See Collaboration mode.
• Acts on what it finds. It can propose a dashboard for the signals you want to keep watching, and it asks for approval before saving a change.
• Reaches your code. Your coding agent can query the same telemetry through Last9 MCP while you work in the repository.

For the investigation method itself, refer to Investigate an incident with AI.

Where Last9 AI works

Last9 AI meets you where the work already happens. It runs in the Last9 dashboard, in the exception and dashboard panels you already open, in your Slack incident channel, and in your coding agent.

In Last9

In your own tools

MCP supplies the tools that query Last9, and skills supply the instructions for using those tools. Your MCP client runs the agent and controls its tool permissions. The managed surfaces in the dashboard and Slack run the agent for you, although the available workflows differ between surfaces.

Enable Last9 AI

You need a Last9 account, AI access for your organization, and relevant telemetry flowing into Last9.

An organization admin can open AI Settings and manage Enable Last9 AI. The same page accepts your own model provider key, so your organization can run Last9 AI on an Anthropic or OpenAI key that it supplies and controls.

If the feature is unavailable in your organization, contact support@last9.io.

After you enable Last9 AI, an admin can give it context about your own systems. Register your remote MCP servers first, then add knowledge that binds their tools.

Privacy and security

AI investigations use your question, conversation context, and tool results to produce an answer. Tool results can contain telemetry such as log messages, span attributes, and exception details. Consider what your telemetry contains when enabling AI access or connecting an external agent.

For MCP, review the connected client’s model provider, data handling, and tool-approval settings. See MCP setup for authentication and agent integration for programmatic access.

Contact support@last9.io for organization-specific AI availability, model configuration, or data-handling requirements.

---

Troubleshooting

• Last9 AI is unavailable in your organization

  Ask an organization admin to check Enable Last9 AI in AI Settings, and to confirm that a model provider key is configured.

• A surface behaves differently from another

  Available workflows and tool permissions differ between surfaces. Check the page for the surface you are using before comparing results.

• You need help with one surface

  See AI Assistant troubleshooting, Slack troubleshooting, or MCP troubleshooting.

Please get in touch with us on Discord or Email if you have any questions.