Overview
Each hosted MCP server exposes a server-specific research surface plus shared helper tools. The helpers cover discovery, health checks, result handling, artifacts, environment inspection, and optional table/database introspection.
| Tool | Description |
|---|
agent_docs | Returns structured MCP documentation for agents: overview, tools, environment, database notes, and examples. |
runtime_health | Returns runtime health and readiness diagnostics. |
deployment_info | Returns deployed URL, runtime profile, storage, and database configuration. |
env_describe | Shows required and optional environment variables with present/missing status and masked previews. |
db_list_tables | Read-only table/view introspection when database access is configured. |
result_preview | Previews a stored result without transferring the full payload. |
result_get | Reads a stored result when the full payload is needed. |
save_tool_result | Persists a stored or raw payload as a scoped downloadable artifact. |
artifact_get_meta | Fetches metadata for a scoped artifact. |
artifact_read_chunk | Reads a scoped artifact in base64 chunks for MCP-only clients. |
artifact_get_download_url | Builds a download URL for a scoped artifact when URL generation is enabled. |
Result and Artifact Flow
Large tabular or nested outputs may be compacted into a stored result_id. Inspect the result with result_preview or result_get. To make a stored or raw payload downloadable, call save_tool_result, then use artifact_get_meta, artifact_read_chunk, or artifact_get_download_url.
{
"request": {
"result_id": "<result-id>",
"filename": "factor_output.json"
}
}
Health and Deployment Discovery
Use runtime_health before long runs and deployment_info when you need to confirm the effective MCP URL, runtime profile, storage settings, or database configuration.
Admin/auth mutation tools and raw database query tools may exist in some deployments but are operator tools, not default client workflow tools.
