Loading ToolBench
A WebMCP-native live auction house where agents bid inside human-set mandates they cannot exceed.
OpenFloor defines 10 tools with clear verb-noun naming (get_*, place_*, set_*, request_*) and consistent descriptions (100-200 chars). All tools have input schemas with typed parameters. However, parameter descriptions are sparse or missing for several tools, and output schemas are not documented. Error handling guidance is absent. The server demonstrates solid foundational structure but lacks the depth expected for production-grade agent tooling.
Perform a dry-run check of whether a bid amount would be accepted, without actually placing it. Returns verdict and any applicable messages about ceiling or approval requirements.
Get the current state of the auction including lot details, current price, high bidder, reserve status, time remaining, and bid count.
Get recent bid history on the current lot, showing bid amounts, bidder aliases, and whether bids were placed by agent or human.
Get detailed information about the current lot including title, condition, seller's estimate, opening price, bid increment, and description.
Get the bidder's session activity including their mandate limits, committed amount, bids placed, and notable events like ceiling changes or confirmations.
Get the bidder's current mandate including hard ceiling, confirmation threshold, headroom to ceiling, supervised band status, automatic bidding setting, strategy notes, and expiration time.
Output schemas not documented. Tools return data but LLMs cannot plan downstream calls without knowing response structure (e.g., what fields does get_auction_state return? Is there a next_cursor for pagination?).
No error handling guidance. Tools lack recovery hints (e.g., 'Bid rejected: amount below minimum. Call check_bid() first to validate.'). Agents cannot self-correct on failures.
Parameter descriptions missing or minimal for read-only tools. get_auction_state, get_lot_details, get_my_activity, get_my_mandate have empty required arrays but no explanation of what context they read (current lot? session state?).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 52 | 2026-07-28+ | v2 |
Place a bid on the current lot. May require human confirmation if the bid exceeds the notify_above threshold. Returns confirmation of placement or explanation of rejection.
Request that the human raise the bidding ceiling. This is a request only—the ceiling remains unchanged until the human approves. Returns confirmation that the request was sent.
Set or update the bidder's mandate with a hard ceiling, confirmation threshold, optional strategy notes, and automatic bidding preference. Validates that confirm-above is not higher than ceiling.
Withdraw from the current lot if the bidder is the high bidder. Returns confirmation or explanation of why withdrawal is not possible.
No pagination support documented. get_bid_history accepts a limit parameter but no cursor/offset or total_count guidance. Large result sets could exhaust context.
Confirmation/dry-run pattern missing for destructive operations. place_bid and withdraw_from_lot are reversible but lack a dry-run or explicit confirmation step to prevent agent mistakes.