Vendor: Cyan Eclipse Technologies · Platforms: MCP over stdio (HTTP transport on roadmap)
Certification scope
CERTIFIED(local) at the NARROWED OBSERVED SCOPE (CLAIM-SCOPE.md): Positive DECLARED, OBSERVED, and INFERRED reports only within CLAIM-SCOPE.md, with evidence or basis carried in the output; OBSERVED is not exhaustive. Withdrawn by name — W1 JS_TS_MULTILINE_BLOCK_COMMENT_CONTINUATION_AFTER_SPLIT_NO_CLOSE_SPAN · W2 JS_TS_CHAINED_NO_CLOSING_SLASH_REGEX_SPANS_ON_ONE_LINE · W3 B_LR_14_INVALID_LINE_NONINVENTORY_IDENTIFIER_FOLLOW · W4 EXHAUSTIVE_OBSERVED_RECALL · W5 DYNAMIC_IMPORT_OBSERVATION_OR_SPECIFIC_UNKNOWN · W6 MTS_CTS_OBSERVED_EXTRACTION · W7 PERMISSION_DENIED_SUBTREE_COMPLETENESS · W8 SELF_REFERENTIAL_SYMLINK_REPORT_AND_SKIP · W9 HOSTED_AUTHENTICATED_OR_MULTI_TENANT_OPERATION. Scoped re-bench card -11 (2026-10-26): 30 pass, 0 na of 30 checks.
What it does
- Read-only static recon of a local codebase: architecture map (languages, packages, entry points, layering signals, file/size distribution)
- Dependency + end-of-life risk list: manifest parsing plus an explicit versioned-in-code EOL rules table — every EOL claim cites its rule id
- Modernization estimate INPUTS (LOC by language, test presence/ratio, manifest age signals, coupling hotspots by fan-in) — raw inputs, never an estimate
- Evidence-bound recon at the NARROWED OBSERVED SCOPE (CLAIM-SCOPE.md): Positive DECLARED, OBSERVED, and INFERRED reports only within CLAIM-SCOPE.md, with evidence or basis carried in the output; OBSERVED is not exhaustive. The three lists are never blended, and anything not statically determinable reports UNKNOWN with the reason
- Safety: target code is never executed or imported, symlinks are never traversed (escapes are reported), binary files are skipped and counted, and size/file-count caps surface as named truncation flags
Tools
recon_scanarchitecture_mapdependency_risk_listmodernization_inputs
Problems it addresses
- Map the architecture of a legacy codebase an agent can analyze
- List dependency and end-of-life risks in an old codebase
- Assess a codebase before a modernization project
Not tested
- Manifest ecosystems beyond the v0.1 parsers listed in README (no Gradle dependency extraction, no lockfile parsing)
- Hosted / multi-tenant operation and auth beyond the stdio process boundary (withdrawn W9)
- Withdrawn W1: JS/TS multiline block-comment continuation after a split no-closing-slash regex-like span (card -10 X1)
- Withdrawn W2: JS/TS chained no-closing-slash regex-like spans on one physical line (card -10 X1)
- Withdrawn W3: B-LR-14 non-inventory identifier follow on invalid JS/TS lines only
- Withdrawn W4: exhaustive OBSERVED recall — B-LR-5 (JS), B-LR-6 (Python), and B-LR-15 / template-${} over-masking recall limits
- Withdrawn W5: dynamic-import OBSERVED entries and a dedicated dynamic-import UNKNOWN (B-LR-3)
- Withdrawn W6: .mts/.cts OBSERVED extraction (suffixes absent from EXT_LANG)
- Withdrawn W7/W8: completeness over permission-denied subtrees and report-and-skip treatment of self-referential symlink loops (B-LR-7)
- B-LR-12 remains a cosmetic bench-suite wording item, not a product claim
Bench scorecard: test-bench/results/eclipse-legacy-recon-mcp-2026-10-26-11.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 →