Local credential safety for AI-assisted workflows with an encrypted vault, scoped injection, output redaction, MCP proxy, and agent guidance.
Hetzer exposes 26 tools with minimal schema rigor. Most tools lack input parameter descriptions entirely (e.g., create_module_recipe, configure_mcp, run_install_wizard have empty {} schemas with no guidance). Tool descriptions are present but generic and often under 50 chars, failing to explain WHEN to use each tool or what it returns. No output schemas documented. Critical security tools (reveal_credential, set_credential, authorize_credential_reveal) lack parameter validation guidance. No error handling patterns visible. Tool naming is reasonable (verb_noun style) but composition is poor, credential management spans 8 separate tools that could be unified. Schemas are either empty objects or minimally typed. This server prioritizes functionality over LLM-usability.
Approve a pending credential request
Audit agent context for security issues
Authorize the revelation of a credential
Configure MCP server settings and parameters
Create a credential request for approval
Create a new module recipe for Hetzer
Create a module recipe from source code or documentation
Diagnose MCP service health and connectivity
Empty input schemas on 8 tools (configure_mcp, create_module_recipe, run_install_wizard, load_module_registry, validate_all_modules, read_audit_events, verify_audit_ledger, diagnose_mcp_service). LLMs cannot infer required parameters or constraints.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., what does list_credentials return? Is it an array of {id, name}? Does it include metadata?). Breaks tool chaining.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 42 | 2026-07-28+ | v2 |
Retrieve a credential request by ID
List all stored credentials in the vault
List all available MCP tools from the catalog
Load the module registry with enabled/disabled modules
Prompt user for a secret value
Read audit log events from the vault
Redact secrets from text and store them in the vault
Resolve module profiles for deployment
Restore redacted secrets in text from the vault
Retrieve and decrypt a credential from the vault
Run the interactive module installation wizard
Execute an MCP tool command with specified arguments
Scan text for leaked secrets and credentials
Store a credential in the encrypted vault
Enable or disable a module
Validate all modules in the registry
Validate a module recipe
Verify the integrity of the audit ledger
Credential tools (reveal_credential, set_credential, authorize_credential_reveal) lack parameter validation guidance. No description of what makes a valid credentialId, what format the value must be, or what happens on auth failure. Invites hallucinated inputs.
Tool 'redact_and_vault' uses 'and' in name, signaling multiple responsibilities (redact + vault). Should split into redact_secrets and vault_secrets so agent can compose independently.
No error handling patterns visible. Tools lack recovery guidance (e.g., 'If credential not found, try list_credentials() first'). No categorization of retryable vs fatal errors. Agents cannot self-correct on failure.