docs/ai-agents/mcp-server/troubleshooting.md
This page covers common issues you may encounter when using the Agent Engine MCP server.
Your agent can't discover the MCP server's tools.
The MCP server fails to start or returns credential errors.
.env file is in the same directory where you run agent-engine commands..env file match the ${env.VAR_NAME} placeholders in your YAML configuration..env values.Authentication failed.
.env file.uv run agent-engine login <organization-id> to re-enter your Client ID and Secret.Authentication succeeded but the request is forbidden.
For list and search results, the MCP server truncates text fields longer than 200 characters to reduce token usage. Truncated fields are marked with [truncated]. To get the full value:
get action with the record ID.get is not available, retry the original query with skip_truncation=true and tight select_fields and limit.If search returns no results but you expect data:
search action only works in hosted mode with the context store enabled. In open source mode, use list with date boundary parameters instead.list with date boundary parameters instead.fuzzy matching instead of like in your filter (e.g., {"fuzzy": {"name": "search term"}}).When querying date ranges that include today, the search index may not have the latest records. Issue both a search call and a list call with date boundary parameters, then merge results and deduplicate by ID.