You added a tool. The new tool works. And your agent quietly got worse at the old ones — calling search_tickets for a refund question meant for search_orders, answering with total confidence, no exception, no alert.
This service is the recovery loop for that failure: register your catalog once and it holds the green state; every loop run measures your live catalog against it per selector class, and when a tool is stealing another tool's calls you get a worklist of disambiguation clauses ranked by measured recovery in re-probe — because a clause that reads well can recover nothing (negation clauses recovered 0.0pp on a token selector) while a sharpening clause mined from the flipped probes recovered the full drop. You pick a fix, the service returns the revised catalog, and a verification re-probe issues a signed recovery receipt with the pair's confusion before and after. A fix that does not measurably recover is receipted NOT_RECOVERED, never RECOVERED.
Prices: Per loop run $0.05 (x402) · Operator $9/mo (up to 200 runs/mo) · Fleet $29/mo (up to 1,000 runs/mo). First loop run free, one time per operator. Small-class probing included; frontier-class probing BYOK or in suites up to 30 probes on company keys.
API: https://textforagents.com/products/tool-confusability-regression-gate/ · Docs: https://textforagents.com/products/tool-confusability-regression-gate/docs · OpenAPI: https://textforagents.com/products/tool-confusability-regression-gate/openapi.json · Support: support@textforagents.com