Loading ToolBench
WebMCP abstraction layer for government websites — install adapters for any state or federal service and let AI agents navigate forms, check eligibility, and submit applications on behalf of citizens.
Civic-MCP exhibits mixed quality across 12 tools. Strengths: all tools have descriptions (10 - 200 chars, within baseline); tool names follow verb_noun pattern (identity_get_profile, check_eligibility, start_application); toolAnnotations present. Only 2 tools (start_withholding_estimate, check_medicare_eligibility) have documented parameters with types and descriptions. No output schemas documented for any tool. Error handling absent, no recovery guidance, no categorization. Composition issues: duplicate tool names (start_application appears twice across adapters; check_application_status vs get_case_status are semantically similar). Parameter descriptions missing entirely for 10 tools. This is a STDIO-only server, capping protocol readiness at 50.
Check the status of Colorado PEAK applications and benefits. Requires a PEAK account: the browser pauses for login; the session is then remembered in the active identity.
Open the BenefitsCal "Do I Qualify" screener for California benefits (CalFresh, CalWORKs, Medi-Cal) and report what it asks. No login required. The screener is interactive — this tool reports the site's questions and content rather than answering on its own.
Estimate when someone becomes eligible for Medicare and what their premium may be, using the official Medicare.gov eligibility calculator. Needs date of birth and whether they worked 10+ years paying Medicare taxes. No login required. Reports what the calculator says.
Open Colorado PEAK's "Find benefits" directory — descriptions of SNAP, Medicaid (Health First Colorado), Colorado Works/TANF, CHP+, and other programs — and report what the site says. No login required.
Get current information about applying for CalFresh (SNAP) in California from getcalfresh.org — what to expect, how applying works — plus the official application URL. GetCalFresh no longer hosts applications itself; applying happens on BenefitsCal (see the gov.california.benefitscal adapter).
10 of 12 tools have empty input schemas ({}). Violates pattern:constrained-input and pattern:tool-description. LLMs cannot infer parameter types or constraints.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields. Violates pattern:tool and pattern:response-shaper.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | 2026-07-28+ | v2 |
Check the status of existing BenefitsCal cases (CalFresh, CalWORKs, Medi-Cal). Requires a BenefitsCal account: the browser pauses so the user can log in; the session is then remembered in the active identity.
Read the applicant profile saved in the active civic-mcp identity. Adapters use this data to prefill government forms. Returns null if no profile has been saved yet.
Save (replace) the applicant profile for the active civic-mcp identity.
Open Colorado PEAK's interactive benefits finder (category checkboxes for food, health, cash, housing, child care, …) and report what it asks. No login required. The finder is interactive — this tool reports the site's questions rather than answering on its own.
Open the Colorado PEAK application for state benefits (SNAP, Health First Colorado, Colorado Works, CHP+) as a guest and hand the browser to the user. Nothing is filled or submitted by the agent.
Open the official BenefitsCal application for California benefits (CalFresh, CalWORKs, Medi-Cal) and hand the browser to the user. This is the same URL GetCalFresh routes applicants to. Nothing is filled or submitted by the agent.
Start the official IRS Tax Withholding Estimator: completes the "About you" step (filing status, age 65+, blind, dependents) and reports what the Income step asks for, so the user can gather pay stubs and figures. Read-only against IRS systems — nothing is filed. No login required.
Duplicate tool names: 'start_application' appears in both gov.california.benefitscal and gov.colorado.peak adapters. LLMs cannot disambiguate. Violates pattern:tool (one thing per tool, clear naming).
No error handling or recovery guidance. Tools that interact with government websites (login required, interactive forms) lack error categorization (retryable vs user-fixable vs fatal). Violates pattern:recovery-guide.
Semantic duplication: 'get_case_status' (BenefitsCal) and 'check_application_status' (PEAK) do the same thing for different states. No unified interface. Violates pattern:tool (avoid multiple tools doing same thing differently).