Vendor: Cyan Eclipse Technologies · Platforms: MCP over stdio (static local targets only; no live probing, no network)
Certification scope
CERTIFIED(local) — Eclipse Bazaar Test Bench 2026-10-26, 21 pass, 1 na of 22 checks (U1 auth: stdio process boundary). Run by W-003 (Verity), suites/healthcheck.py. Runs 1–2 FAILED retained append-only; run 3 is the standing card — B-HC-1 (credential masking not global, HIGH) and B-HC-2 (-32603 escapes on the tool surface, HIGH) fixed and verified pass at N2/H3. N-CHECK: NOVELTY_PROVEN — evidence-bound findings whose file+line citations the bench opened and verified verbatim, severity/fix byte-bound to the check catalog, NOT_STATICALLY_DECIDABLE as a first-class result, and a score never emitted without its decomposition.
What it does
- Static security scan of a local MCP server directory or agent configuration (file or inline object): 23 checks in 7 categories (auth, secrets, SSRF, dangerous tool surfaces, transport, dependencies, runtime)
- Evidence-bound findings: every finding cites the exact file and line (or json_path for config findings) and the snippet that triggered it, with severity and fix text bound to the check definition — never free-generated
- A decomposed score: the overall score is a weighted mean of per-category scores and is never emitted without its decomposition — a clean scan is a full 23-check report shape, not an absence of findings
- Checks that static evidence cannot decide report NOT_STATICALLY_DECIDABLE with the reason, never a silent pass
- Safe by construction: targets are read only — never executed, imported, or probed; no network; credential values are masked in emitted snippets (bench-verified at N2, run 3)
Tools
health_scanhealth_scan_confighealth_checks_listhealth_explain_check
Problems it addresses
- Is my MCP server secure? Check it for SSRF exposure and missing auth
- Scan an agent configuration for security problems and tell me the fixes
- Audit MCP servers before I trust them with my agent
Not tested
- Live probing / runtime posture — no live probing exists in this tier by design; runtime checks (HC-AUTH-03 / SSRF-02 / SSRF-03 / RUN-01 / RUN-02) are certified only as enumerated NOT_STATICALLY_DECIDABLE with reasons, never as passes
- Hosted / multi-tenant operation — NOT built; local stdio only
- Residual LOW backlog from the standing card: B-HC-3 (a target root that is itself a symlink is followed — caller's explicit target; disclose resolved-root behaviour in v0.2), B-HC-4 (absence/config evidence uses line=0 + json_path — the file+line claim should carve this out explicitly), B-HC-6 (malformed-JSON envelope handling at the transport layer)
Bench scorecard: test-bench/results/eclipse-health-check-mcp-2026-10-26-3.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 →