Loading ToolBench
Find what a role actually pays, including where the employer publishes nothing -- reads public job-board APIs (Greenhouse, Ashby, Lever, SmartRecruiters, Workday) for published salary bands, per seniority and per currency.
payband-mcp has 7 tools with complete input schemas and descriptions. Naming follows verb_noun convention (company_topics, company_posts, read_post, find, research, job_openings, pay_bands). Descriptions are substantive (100-250 chars), explaining WHAT each tool does and WHEN to use it. However, output schemas are not documented, LLMs cannot predict return structure. Parameters lack type constraints (enums, ranges, patterns). Error handling is absent from visible code. Five Blind-backed tools are conditionally disabled via ENABLE_BLIND flag, which is a pragmatic workaround but reduces discoverability. No tool annotations (readOnlyHint, destructiveHint) despite all being READ_ONLY. Composition is sound, each tool has one responsibility and tools chain naturally (company_topics → company_posts → read_post).
List posts about a company, optionally narrowed to one topic. `topic` accepts a bare keyword from company_topics (e.g. "india", "wlb").
List the discussion topics Blind itself suggests for a company. These are the highest-signal entry points -- e.g. Roku exposes india, wlb, culture, layoffs, interview, rsu. Use one as the `topic` for company_posts.
Search Blind for posts matching a question about a company. Returns the most relevant posts, ranked by overlap with the question and by upvotes. Noise (low-content replies, referral spam) is filtered out before ranking.
List open positions at a company across all public job boards. Searches Greenhouse, Ashby, Lever, SmartRecruiters, and Workday for postings matching the company name and optional role keyword.
Published salary bands for a role at a company, by seniority and market. Aggregates pay ranges from all public job boards, groups by seniority level, and reports the typical band, spread, and how many postings share each band.
Output schemas not documented. LLMs cannot predict return structure (fields, types, pagination). Baseline: 100% of A+ tools document return types.
Parameters lack type constraints (enums, ranges, patterns). E.g., 'topic' accepts bare keywords but no enum list; 'limit' and 'page' are unbounded integers. Baseline: constrained inputs prevent hallucinated values.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All 7 tools are READ_ONLY but this is not declared in schema. Agents cannot infer safety properties.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | 2026-07-28+ | v2 |
Read one Blind post in full. Returns the body, Blind's own AI summary of the comment thread, and the comments with each commenter's employer -- which is how you weigh a claim (an answer from someone at the company differs from a passer-by).
Research a company on Blind, returning the most relevant posts and comments. Combines find and read_post to return a curated set of posts with their top comments, ranked by relevance to the question.
Error handling not visible in source. No recovery guidance, error classification, or actionable error messages. Pattern: recovery-guide.
Five Blind-backed tools conditionally disabled via ENABLE_BLIND flag. When disabled, tools are not registered, reducing discoverability. Agents cannot see these tools exist until flag is set.