Loading ToolBench
Federal Audit Clearinghouse (FAC) MCP — US single-audit data. Provides access to US single audits (Uniform Guidance / former OMB A-133) filed with the Federal Audit Clearinghouse, enabling queries by auditee name, UEI, EIN, state, audit year, and federal dollars expended.
Five tools with complete input schemas and strong descriptions (avg 180 chars). All parameters are typed and described. Tool names follow verb_noun pattern (fac_search_*, fac_get_*, fac_audit_*). No output schemas documented, a significant gap for LLM planning. Error handling and recovery guidance absent. Parameters accept aliases (e.g., 'query' for 'auditee_name') which is good for natural identifiers, but no enum constraints on free-form fields like 'auditee_name' that could hallucinate invalid values. No security annotations or permission declarations.
Audit findings recorded against single audits — the compliance exceptions, material weaknesses, significant deficiencies, questioned costs and repeat findings that auditors reported. Query by report_id, or by recipient name/UEI to sweep every audit that recipient filed. Answers "has this grant recipient had audit findings, and were they material or repeated".
List the organizations that expended money under a given federal Assistance Listing / CFDA number (e.g. "93.224" community health centers, "84.010" Title I, "20.205" highway planning), ranked by dollars expended, with the recipient name, state, audit year and whether it was a major program. Answers "who spends the money in this federal grant program and how much does each one run".
Fetch one single audit by its FAC report_id, enriched with the full federal program detail: every Assistance Listing (CFDA) number the entity expended money under, the program name, dollars expended per program, whether it was audited as a major program, and the opinion type. Also returns auditee contact and auditor details, plus audit findings when the FAC findings table carries them. Answers "what did this specific audit cover and what did it conclude".
Year-by-year single-audit history for one grant recipient, identified by UEI or by name. Returns each audit year with total federal awards expended, the report_id, the oversight agency and the FAC acceptance date, plus a growth read on the spend trend. Answers "how has this organization's federal funding moved over time and which years were audited".
No output schemas documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., report_id for chaining). Responses must include structured field definitions.
No error handling or recovery guidance. API errors (e.g., rate limits, not found) return raw responses without actionable next steps for the LLM.
Free-form string parameters (auditee_name, cfda) lack enum constraints or validation rules. LLMs may hallucinate invalid values. Descriptions mention formats (e.g., '93.224' or '93-224') but no formal pattern constraint.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2026-07-28+ | v2 |
Search US single audits (Uniform Guidance / former OMB A-133) filed with the Federal Audit Clearinghouse by auditee name, UEI, EIN, state, audit year, or minimum federal dollars expended. Returns the report_id, the organization name and location, total federal awards expended for the year, the cognizant/oversight agency, and the date FAC accepted the submission. Answers "does this grant recipient file a single audit and how much federal money does it run".
No tool annotations (readOnlyHint, idempotentHint). All tools are read-only, but this is not declared in the schema, forcing LLMs to infer safety.
Pagination parameters (limit, offset) present but no total count or next_cursor returned. Large result sets risk context window exhaustion without clear continuation guidance.