MCPg v0.6.12 — release notes

Released: 2026-07-28 Tool surface: 254 tools across 19 capability buckets (read-only mode exposes a subset) Tests: unit + integration suite green (PG 14 / 15 / 16 / 17 / 18 / 19 / WarehousePG) Runtime: Python 3.14

A 0.6.11 → 0.6.12 release that adds two tools — a long-running analytical read path and a cache-freshness escape hatch — plus a dependency security bump. Backward-compatible: no existing tool signature changed, and both new tools default to the previous behaviour being available (the analytical tool is on by default; the cache still behaves exactly as before unless you ask for fresh data).

New: run_analytical_query — long-running reads on an isolated pool

run_select is deliberately bounded to a short (~30 s) timeout on the shared connection pool so an agent can’t pin a connection with a runaway query — which made genuine analytical queries (large aggregations, multi-table joins, window functions) infeasible, and MCPG_STATEMENT_TIMEOUT_MS didn’t help (it moves Postgres’s statement_timeout, not the client-side asyncio cap).

The new tool runs a read-only SELECT through the same allowlist + tenancy/RLS + read-only transaction as run_select, but on a dedicated connection pool isolated from the main one, with an elevated, bounded timeout — so a slow analytical query can never starve the fast-path tools. A per-call timeout_ms (clamped to the configured maximum) and optional work_mem are exposed, and a run_select timeout now points the agent at the tool so it needn’t predict a query’s duration up front.

Boot-time knobs: MCPG_ENABLE_ANALYTICAL_QUERIES (default true — the authoritative off-switch; set false to withdraw the tool), MCPG_ANALYTICAL_TIMEOUT_MS (default 120000, 2 min), MCPG_ANALYTICAL_MAX_TIMEOUT_MS (600000, 10 min ceiling), and MCPG_ANALYTICAL_MAX_CONCURRENCY (2 — the isolated pool size, which is also the concurrency cap). Read-only and primary-database only for now.

New: cache-freshness controls for out-of-band schema changes

MCPg’s read cache is invalidated automatically by MCPg’s own write/DDL tools, but it could serve stale introspection/advisor results for up to MCPG_CACHE_TTL_SECONDS (default 300 s) after a schema change made outside MCPg — a direct psql/migration change, another connection, or a second process. Re-running index/constraint validation after altering a foreign key could return the pre-change answer. Two escape hatches:

• a per-call fresh: bool = False argument on the introspection/advisor reads (describe_table, list_indexes, list_constraints, list_foreign_keys, get_compact_schema, recommend_indexes, audit_database) — bypasses the cache read, re-queries live, and refreshes the entry;
• a new clear_cache tool that flushes the whole result cache in one call.

Security: mcp SDK bumped to ≥ 1.28.1

Raised the mcp SDK floor from ≥ 1.25.0 to ≥ 1.28.1 to clear three advisories in the resolved mcp 1.27.1 (CVE-2026-52870, CVE-2026-52869, CVE-2026-59950). No MCPg source change was required — the per-request tenancy role path is unchanged in 1.28.1, and the full unit + contract suite passes. pip-audit --strict on the resolved runtime deps is clean.

Also in this release

• run_analytical_query hardening. Query timeouts raise a typed QueryTimeoutError detected structurally across the full __cause__/__context__ exception chain — covering both the client-side asyncio cap and the server-side statement_timeout (SQLSTATE 57014) — so the self-correcting hint fires reliably and locale-independently.
• pglast 7.15 → 8.x (the SQL-safety kernel’s parser), tracking the PostgreSQL 18 grammar. No behaviour change — the full adversarial + fuzz + differential-parity SQL-kernel suites pass unchanged.

Upgrade

pip install --upgrade mcpg
docker pull ghcr.io/devopam/mcpg:0.6.12   # or :latest

Or grab mcpg-0.6.12.mcpb from this release and double-click it into Claude Desktop. No configuration changes required.

Full changelog

See ../CHANGELOG.md [0.6.12] for the complete itemised list.