What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
You can connect a Copilot Studio agent to a Litmus MCP server so it can discover chaos-engineering capabilities and, if permitted, invoke them. The key is to identify the exact Litmus product and server first: the available repositories describe two different product families, and their tools and operations are not interchangeable. Copilot Studio connects to an MCP server; it does not make an incompatible endpoint or an overly broad set of experiment permissions safe.
First identify which Litmus system you are connecting
“Litmus MCP server” can refer to more than one implementation. Before configuring Copilot Studio, check the repository owner, product, endpoint, tool catalog, credential requirements, and transport for the server you actually plan to run.
As an Amazon Associate I earn from qualifying purchases.
| Repository | Product and API surface described | What to verify |
|---|---|---|
| litmuschaos/litmus-mcp-server | LitmusChaos 3.x Chaos Center operations, including experiments, infrastructure, environments, experiment runs, probes, and ChaosHub. | Current release, exact tool names and inputs, supported transport, endpoint, and authentication for your deployment. |
| litmusautomation/litmus-mcp-server | Litmus Edge and Edge Manager operations, including device, telemetry, and management capabilities. | Confirm that you intend to manage Litmus Edge rather than run LitmusChaos experiments; then verify the server’s current tools, transport, endpoint, and authentication. |
Litmus describes its platform as an open-source chaos-engineering platform with declarative experiments, steady-state probes, and Prometheus-based observability. Its product description of an MCP server covers discovering, running, and monitoring experiments. Those capabilities do not establish that every MCP server bearing the Litmus name offers the same API or that a particular Copilot connection has been tested.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsStart with the LitmusChaos project site for the platform context, then use the README and release information for the exact repository and version you deploy.
#1 Best Overall
Check transport and endpoint compatibility
Copilot Studio supports Streamable MCP transport. Microsoft says its support for MCP over SSE ended after August 2025, so an SSE-only endpoint is not the right choice for a new Copilot Studio connection. The Litmus Automation repository documents http://<host>:8000/mcp as its Streamable HTTP endpoint and /sse as legacy. That endpoint detail applies to that repository’s documentation; do not assume it is the endpoint for the separate LitmusChaos server.
- Confirm that the selected server exposes Streamable MCP and identify its actual URL.
- For a production deployment, use HTTPS and verify the endpoint is reachable from the Power Platform environment.
- Check that the server’s authentication method is supported by your chosen Copilot Studio connection path.
- Do not treat a sample client configuration as proof of compatibility. The Litmus Automation README’s VS Code/GitHub Copilot example uses SSE and notes that clients supporting Streamable HTTP should use
/mcpinstead.
Choose how to add the server to Copilot Studio
Microsoft documents two direct ways to connect an existing MCP server: the in-product MCP onboarding wizard, which it recommends, and a Power Apps custom connector. A registered and approved BYO MCP server through Agents 365 is another documented route, but availability can depend on preview status and tenant configuration.
| Connection route | Setup and schema | Authentication and sharing considerations |
|---|---|---|
| MCP onboarding wizard | Use the agent’s Tools page, then Add a tool > New tool > Model Context Protocol. Enter the server name, description, and URL. The server supplies its MCP tool definitions. | Choose no authentication, API key, or OAuth 2.0 according to the server and tenant setup. This is Microsoft’s recommended path for an existing MCP server. |
| Power Apps custom connector | Create a custom connector using an OpenAPI YAML schema, create its connection, and then add the server or connector to the agent. | Requires maintaining the connector schema and configuring authentication for the connector and target server. Consider this route when a custom connector is needed for your deployment or sharing model. |
| Agents 365 approved BYO server | Register and approve the MCP server through the applicable Agents 365 workflow, then connect it to the agent. | Check current preview and tenant availability before designing around this route. |
Both direct setup paths depend on the selected server’s actual protocol and authentication support. Microsoft’s connection guidance is at Connect your agent to an existing MCP server.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Wizard setup sequence
- Open the agent in Copilot Studio and go to Tools.
- Select Add a tool, then New tool, then Model Context Protocol.
- Enter a recognizable server name, a description that identifies its product and scope, and the verified Streamable MCP URL.
- Select the authentication option that matches the server: no authentication, API key, or OAuth 2.0. Configure any required credentials using your organization’s approved secret-handling process.
- Complete the connection and review the tools Copilot Studio imports from the server.
Enable orchestration and keep the tool scope deliberate
Generative orchestration must be enabled for MCP use. Copilot Studio receives each MCP tool’s name, description, inputs, and outputs; these details inform how the agent selects tools in response to a request. Review the imported definitions for clarity and scope before exposing the agent to users.
Rank #3
Copilot Studio initially allows all tools from a connected server. You can disable individual tools. If you switch off Allow all, tools subsequently added to that server default to off, so review them deliberately when the server changes. Microsoft also documents that an MCP server owner must configure a resource as an output of one of its tools for the agent to use that resource.
- Keep discovery and status tools available only where they serve the agent’s purpose.
- Disable experiment execution or other write operations unless the use case requires them.
- Revisit the enabled-tool list after server updates; a new tool should not silently expand the agent’s intended scope.
- Review the Power Platform data policy that applies to the connector and tenant.
See Microsoft’s guidance on adding MCP tools and resources to an agent and extending an agent with MCP.
Plan permissions and approvals for disruptive actions
An agent that can launch experiments is an operational interface, not just a chat surface. Microsoft states that the connecting party is responsible for the external MCP server and its tools and resources. Apply controls in the server, credentials, environment, and organizational change process rather than assuming the agent will enforce them.
- Use least privilege. Limit credentials to the intended projects, namespaces, infrastructure, and environments. Avoid credentials that can affect unrelated or production systems.
- Separate read and write capability. Where the server allows it, expose discovery and monitoring separately from experiment execution or other changes. Disable write tools the agent does not need.
- Protect credentials. Use the authentication method supported by the selected server and connection route; store and rotate secrets under your organization’s approved process. Do not place credentials in agent instructions or user-facing descriptions.
- Make target selection explicit. Require the request or workflow to identify the intended environment and experiment, and have the agent confirm material parameters before a disruptive operation.
- Use human review where required. Add a human approval or change-control step for higher-risk actions when your organization requires one. This is a deployment recommendation, not a claim that Litmus or Copilot Studio automatically imposes a particular approval mechanism.
Validate with a staged workflow
Test the connection and operational boundaries in a non-production environment before making execution tools available for routine use. Litmus describes experiment discovery, execution, monitoring, probes, and observability as platform capabilities; the sequence below is a proposed validation workflow, not a tested run or a guarantee that every server exposes these exact operations.
Quick Recap
- Discover. Ask the agent to list the experiments or capabilities available from the selected server. Check that its answer matches the server’s tool catalog and does not mix Litmus Edge operations with LitmusChaos experiments.
- Inspect. Have it identify the intended target and the applicable steady-state checks. Confirm the environment, scope, and parameters against the server’s actual tool inputs.
- Run a low-risk exercise. In non-production, request a suitable experiment only after verifying that the selected tool, credentials, and any required human review are in place.
- Monitor. Ask for the run status and probe results using the server’s available monitoring tools. Where configured, use the platform’s observability data to assess whether the expected steady state held.
- Handle interruption or failure using the real catalog. Determine from the selected server’s tool definitions how to stop, investigate, or recover. Do not invent a stop command or assume a tool exists if its schema does not provide it.
Common setup failures to check
- The server cannot be added or reached: check endpoint reachability from the Power Platform environment, HTTPS configuration, and whether the URL belongs to the intended Litmus product and server.
- The connection fails at protocol negotiation: verify that the endpoint is Streamable MCP, not an SSE-only URL. For the Litmus Automation repository, its README distinguishes Streamable HTTP at
/mcpfrom legacy/sse. - Tools do not appear or the agent does not invoke them: check that the connection completed, generative orchestration is enabled, the tools are allowed, and the tool descriptions and schemas are appropriate to the request.
- A resource is unavailable: confirm with the server owner that it is configured as an output of one of the server’s tools.
- An operation is rejected: check the selected server’s authentication, credential permissions, target scope, and tenant connector policies rather than broadening access by default.
- A newly added tool is unavailable: if Allow all is off, inspect the tool settings because newly added server tools default to off.
What to confirm before rollout
- The exact Litmus repository, product family, deployed release, endpoint, tool names, and input schemas.
- Streamable transport and authentication compatibility with the target Copilot Studio tenant and environment.
- The available connection route and any tenant-specific connector or BYO server policies.
- The tools the agent is allowed to use, the environments its credentials can reach, and the human approval controls required for disruptive work.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
