You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Link to the gh-aw PR: github/gh-aw#46919 — MCP gateway startup: degrade on optional server failures, fail only on required/no-success cases
This change introduced a required/optional distinction for MCP server startup failures. Before this fix, any single MCP server returning a non-200 during initialize would hard-fail the entire gateway start. There is no test exercising optional-server degradation behavior.
Configure a workflow with two MCP servers: one that works (e.g., safeoutputs) and one optional server configured with a bad/unreachable URL. The gateway should start successfully, degrading gracefully on the optional server failure. The agent should still be able to create an issue via the working MCP server.
New fixtures or secrets needed
Needs a way to inject a bad MCP server URL in the frontmatter, or use a known-unreachable endpoint. The shared/mcp/ definitions might be extendable. Alternatively, use a local test HTTP server — but that may require new fixture work.
Notes
Open question: can the test frontmatter specify an ad-hoc MCP server with an invalid URL, or only reference shared/mcp/ definitions? If an invalid URL can be inlined, no new fixtures are needed. This tests a defensive path that currently has zero coverage.
Warning
Firewall blocked 1 domain
The following domain was blocked by the firewall during workflow execution:
awmgmcpg
To allow these domains, add them to the network.allowed list in your workflow frontmatter:
Motivation
Link to the gh-aw PR: github/gh-aw#46919 — MCP gateway startup: degrade on optional server failures, fail only on required/no-success cases
This change introduced a required/optional distinction for MCP server startup failures. Before this fix, any single MCP server returning a non-200 during initialize would hard-fail the entire gateway start. There is no test exercising optional-server degradation behavior.
Proposed test
test-copilot-mcp-optional-server-degradation.mdworkflow_dispatchcreate-issueMinimal test prompt sketch
Configure a workflow with two MCP servers: one that works (e.g., safeoutputs) and one optional server configured with a bad/unreachable URL. The gateway should start successfully, degrading gracefully on the optional server failure. The agent should still be able to create an issue via the working MCP server.
New fixtures or secrets needed
Needs a way to inject a bad MCP server URL in the frontmatter, or use a known-unreachable endpoint. The shared/mcp/ definitions might be extendable. Alternatively, use a local test HTTP server — but that may require new fixture work.
Notes
Open question: can the test frontmatter specify an ad-hoc MCP server with an invalid URL, or only reference shared/mcp/ definitions? If an invalid URL can be inlined, no new fixtures are needed. This tests a defensive path that currently has zero coverage.
Warning
Firewall blocked 1 domain
The following domain was blocked by the firewall during workflow execution:
awmgmcpgSee Network Configuration for more information.