Vendor: Cyan Eclipse Technologies · Platforms: MCP over stdio
Certification scope
CERTIFIED(local/simulator) — Eclipse Bazaar Test Bench 2026-10-25, 23/23 checks. Run by W-003 (Verity), suites/airtime.py. The original PARTIAL card (20/23; D1 MEDIUM + D2 HIGH) is retained append-only; the scoped re-bench card -2 closed both defects and is the standing verdict basis. N-CHECK: NOVELTY_PROVEN — spend enforcement happens pre-call, ad copy is bound to a signed claims record at the publish path, and every mutation or refusal seals into an advertiser-held evidence ledger. Simulator/test-double only; no live transport, campaign, or spend was tested.
What it does
- Spend enforcement pre-call: every campaign mutation passes a gate against a Herald-signed mandate (account × platform ceilings, geo/audience bounds, and flight windows) before any platform call; a refusal makes zero platform calls
- Claims binding at the publish path: copy may publish only from a signed claims record, with paraphrase screening, special-category classification, implied-attribute screening, crypto permission classes, and political verification as versioned policy data
- Advertiser-held evidence that survives the account: mutations and refusals seal into an append-only, hash-chained, signed ledger and export as an offline-verifiable appeal pack
- A closed refusal vocabulary and no override tool at any tier — more spend requires a new signed mandate, never an exception path
- P0 is Meta-first and local/simulator only: the transport is a simulator/test-double, no live spend exists, and no campaign activates for real
Tools
mandate_issuemandate_getmandate_checkmandate_revokeclaims_record_upsertclaims_record_getcopy_screenad_account_listcampaign_listcampaign_getinsights_getcampaign_create_draftcampaign_updatecampaign_set_statusaudience_attachledger_getappeal_pack_export
Problems it addresses
- An ad platform can execute a campaign change before anyone checks the signed spend mandate
- Ad copy drifts into claims we cannot substantiate — or claims that were retracted
- When an ad account closes, I need evidence and appeal history that survive the account
Not tested
- Live Meta transport — NOT built and out of P0 scope: simulator/test-double only; no live spend exists.
- Standalone Claims Checker as a separate Bazaar Tool — P1 per spec §3.3; P0 ships the internal harness form (tools/claims_checker.py), exercised as part of the benched product surface.
- Google/TikTok/X/LinkedIn/Reddit phases (P1/P3), hosted connector + CASA posture (P2), and Till billing composition (P2).
- Production Herald custody for the signing key — P0 uses a local seed key (keys/, gitignored, chmod 600); production custody is a hard gate before any P2/live scope.
Bench scorecard: test-bench/results/eclipse-airtime-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 →