-
Notifications
You must be signed in to change notification settings - Fork 8
Closed
Description
I'm attempting to configure Claude Code to use the Harness MCP server. To test it, I passed in an execution URL of a failed build and asked Claude to tell me why it failed.
Claude is able to use the get_execution
API without any issues, but unfortunately fails when attempting to call the download_execution_logs
API.
⏺ harness - get_execution (MCP)(plan_execution_id: "<omitted>", org_id: "<omitted>", project_id: "<omitted>")
⎿ {
"pipelineIdentifier": "<omitted>",
"projectIdentifier": "<omitted>",
… +74 lines (ctrl+r to expand)
⏺ The execution shows a failure with multiple "exit status 1" errors, but the details are generic. Let me download the
detailed logs to see the specific failure reasons.
⏺ harness - download_execution_logs (MCP)(plan_execution_id: "<omitted>", logs_directory: "/tmp", org_id:
"<omitted>", project_id: "<omitted>")
⎿ Error: MCP error -32603: failed to fetch log download URL: failed to fetch log download URL: received client side error
status code 403
Here's how I've configured Claude Code via .mcp.json
:
{
"mcpServers": {
"harness": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"HARNESS_API_KEY",
"-e",
"HARNESS_DEFAULT_ORG_ID",
"-e",
"HARNESS_DEFAULT_PROJECT_ID",
"harness/mcp-server",
"stdio"
],
"env": {
"HARNESS_API_KEY": "${HARNESS_API_KEY}",
"HARNESS_DEFAULT_ORG_ID": "${HARNESS_DEFAULT_ORG_ID}",
"HARNESS_DEFAULT_PROJECT_ID": "${HARNESS_DEFAULT_PROJECT_ID}"
}
}
}
}
Metadata
Metadata
Assignees
Labels
No labels