Approved-surface verdict per directory audience — by TextForAgents LLC. Per directory audience, proof that what that audience serves today is what its reviewer approved — and a named verdict when it isn't. A multi-directory MCP publisher serves several audiences at once over one shared codebase, each under different approval semantics: ChatGPT stores the published tool surface as a reviewed snapshot that only a new submission can change, Claude live-reads the current surface on the next connection with no resubmission, and registry/directory listings are hand-maintained metadata that drift separately. This service records the signed per-audience, per-configuration approved reference (name, title, description, inputSchema, annotations), re-serves each audience's surface through its facade and overrides, diffs it field-by-field, and returns intent-separated verdicts — MATCHES-APPROVED / ACCIDENTALLY-DIVERGED / INTENTIONALLY-CHANGED-RESUBMISSION-OWED / SERVING-STALE-APPROVED / LISTING-DRIFT / POLICY-DRIFT-SUSPECT / CAPTURE-TAMPERED — with a post-override canary proving the diff path can fire per configuration. Stated plainly: audience serve paths are faithful harnesses modelling the documented directory semantics (reviewed snapshot / live-read / hand-listed, labelled as such in every verdict record) — real-directory fidelity against the live directories is unmeasured; willingness to pay is unmeasured.
Verdict query $0.01/query · Per-audience surface verdict $0.99/grade · Full multi-audience verdict $4.99/verdict · Publisher watch pass $19.99/server/month (20 audience-grades per server per month included; grades beyond metered at $0.99). One free per-audience verdict per publisher, ever.
Product: https://textforagents.com/products/approved-surface-verdict-per-directory-audience/ · Support: support@textforagents.com