@arizeai/phoenix-mcp

Analyzes a stock based on its ticker symbol and provides investment recommendations.

Optimizes a portfolio based on stocks, risk tolerance, and investment horizon.

Forecasts market trends for a specific sector over the given timeframe (in months).

Analyzes an economic indicator and its impact on financial markets.

Generates a comprehensive investment report based on client profile and market analysis.

List the Phoenix tool groups that can be revealed for this session. Tools are grouped by domain. Only a small default set is visible up front; call ``enable_tool_group`` to reveal a group's tools before using them.

Return the allowlisted analytics SQL schema for telemetry, datasets, and experiments.

Execute read-only analytics SQL against allowlisted Phoenix tables. Returns either the columns, rows, and applied limits, or an error envelope when the SQL cannot be accepted. `row_count_is_partial` is the authoritative answer to whether the result was truncated by either row or response-byte limits. One row beyond the row limit is fetched, and `notes` identifies the limit that applied. `estimated_rows` is available only on PostgreSQL, where it is the planner's untruncated-row estimate. It is not a count, it can be out by a large factor over JSON paths, and it never answers the truncation question -- that is what the flag is for. `applied` describes the effective dialect, row limit, and rewrites; `backend_validated` says whether the backend execution gate ran; `notes` lists caveats callers should not infer. Code-mode `call_tool` already returns this envelope as a dictionary. Check for an `error` key before reading `rows`. Preserve any error in your summary; do not call `json.loads` on the result. With `validate_only=True`, Phoenix validates the statement but does not execute it for data; the successful empty result carries a note saying so.

y

y

y