Vendor: Cyan Eclipse Technologies · Platforms: MCP over stdio (HTTP transport on roadmap)
Certification scope
CERTIFIED(local) — Eclipse Bazaar Test Bench 2026-10-25, run 2: 16/17 checks pass, 1 na (U1 auth: stdio process boundary). Run 1 FAILED card retained in results/ — both failures were bench-suite assertion bugs (schema key name; hash assertion over schema_read receipts, which take no SQL), fixed in the suite, no product code touched. Run by W-003 (Verity), suites/deepfield.py.
What it does
- Per-database policy with table allowlists: unlisted tables are indistinguishable from nonexistent ones
- Read-only by default; single-statement classifier blocks writes smuggled inside string literals
- EXPLAIN cost gate refuses full scans before execution; indexed reads pass
- Row caps on every read; PII columns masked (hash / partial / null) before results leave the tool
- Writes only inside scoped, capped, expiring mandates: dry-run with before-image, then execute
- Hash-chained receipts for every query — blocked ones carry the SQL hash only, never raw SQL
Tools
db_listschema_readquery_readquery_explainwrite_dryrunwrite_executereceipt_list
Problems it addresses
- I want an agent to query my database without ever seeing tables I did not allowlist
- PII must be masked before query results reach the agent
- Blocked queries belong on the audit record, not just the ones that ran
Not tested
- Postgres LIVE path NOT tested — no Postgres server binary, psql, or Python driver exists on this host; the bench verified only the fail-closed posture (all tools return ERROR verdicts). The adapter is implemented behind a driver interface and was tested by its own suite against a recording-fake transport only. Live verification is spec phase P1 and is NOT claimed.
- MySQL/MariaDB, SQL Server, MongoDB, Redis, warehouse (Snowflake/BigQuery), and vector adapters — roadmap phases P1–P4, not built.
- Automated rollback tool — v0.1 retains before-images under a rollback handle; restore is a documented operator procedure (P1).
Bench scorecard: test-bench/results/eclipse-deepfield-mcp-2026-10-25-2.json
Evaluation-intake listing
This listing records the tool's place in the Bazaar's evaluation
and monitoring pipeline, with the evidence gathered so far printed above — verdict, scope,
and not-tested items included. It is not sold on this site, and its presence in the catalog
is not a claim beyond that evidence. The service behind every verdict is the bench:
what a bench run buys →