MCP Server for Authorised Observability Tools

What Guance MCP Server does

Turn supported Guance capabilities into structured MCP tools for an authorised client

The server provides a standard tool interface; it does not grant unrestricted platform access or make an Agent’s conclusion automatically correct. Tool visibility follows the API Key permissions, and write tools can modify resources. Review the tool, inputs, returned evidence, and approval boundary for each workflow.

Connect clients that support the documented MCP transport

Configure the endpoint, authentication, and tool approval behaviour in the client. Client-specific fields and capabilities can differ, so verify the current Guance and client documentation.

Trae

Cursor

Windsurf

Claude Desktop

VS Code

Codex

What an MCP integration must control

No live context: an assistant cannot verify current system state from conversation text alone.

Broad credentials: an over-permissioned connection can expose data or actions beyond the intended task.

Opaque tool use: a conclusion is difficult to trust when the selected tool, parameters, and result are hidden.

Write risk: management tools can change workspace resources if an Agent runs them without approval.

How Guance provides the tool path

Structured tools: MCP provides descriptions and input schemas that a compatible client can inspect.

Credential boundary: Guance tool visibility and results follow the configured API Key permissions.

Evidence path: queries return data that can be reviewed separately from the model’s narrative.

Approval control: teams can require confirmation before write tools or sensitive workflows run.

From connection to evidence-led investigation

Connect an MCP-compatible client to Guance

Configure the documented endpoint and authentication, verify the connection, and inspect the tool catalog before giving the client an operational task.

View documentation

Connect an MCP-compatible client to Guance

Ground the conversation in returned observability evidence

Query a defined service, resource, or time window and keep the tool inputs and returned records available for engineering review.

View documentation

Ground the conversation in returned observability evidence

Observe the Agent while governing the tools it can call

Use Agent Observability for the Agent execution path and MCP permissions or approvals for access to Guance operations.

View documentation

Observe the Agent while governing the tools it can call

Related AI and observability capabilities

OWL CLI

Use authorised Guance tools from terminals, scripts, and Agent workflows.

AI Agent Observability

Trace the sessions, workflows, model calls, and tools used by an Agent.

Unified observability platform

Connect telemetry, service context, and incident workflows in one operating model.

Guance MCP Server questions

What is Guance MCP Server?

Guance OWL MCP Server exposes supported Guance tools to compatible AI clients through the Model Context Protocol. It lets a client request observability context or supported actions through structured tool calls.

How are data access and write actions controlled?

Visible tools, accessible resources, and execution results follow the configured API Key permissions. Write tools can modify workspace resources, so Guance recommends a manual confirmation or approval mechanism for Agent and MCP-client workflows.

When should I use MCP Server instead of OWL CLI?

Use MCP Server when an MCP-compatible AI client should discover and call Guance tools directly. Use OWL CLI for terminal, script, or command-oriented Agent workflows. Both use the same tool-oriented Guance capability layer.