Type-aware, secret-safe clipboard history exposed to Claude via MCP. macOS + Linux.
Strong foundation with 13 well-named tools using clear verb_noun patterns (list_, get_, search_, copy_, pin_, tag_, delete_, clear_). All tools have descriptions (avg 95 chars, within 10-1024 range). Input schemas are complete with typed parameters and descriptions. However, output schemas are not documented, responses are returned as JSON strings without formal schema declarations. Error handling is minimal; most tools return generic 'error: {}' messages without recovery guidance. No tool annotations (readOnlyHint/destructiveHint) despite clear risk stratification. Destructive tools (delete_item, clear_history) lack confirmation patterns.
Clear clipboard history by scope.
Restore a clip to the system pasteboard.
Hard-delete a clipboard entry and its secrets row if applicable.
Return code clips, optionally filtered by language.
Fetch one clipboard entry by id. For secrets returns metadata only. Pass with_blob=true to inline an image/file blob as a base64 data URL.
Return JSON clips with parsed structure preview.
List secret clip metadata only — no values.
Output schemas not documented. Tools return JSON strings without formal schema declarations. LLMs cannot predict response structure for planning downstream calls.
Destructive tools (delete_item, clear_history) lack confirmation or dry-run patterns. Agents can permanently delete clipboard history without explicit user approval.
Error handling returns generic 'error: {}' strings without recovery guidance. LLMs cannot determine if errors are retryable, user-fixable, or fatal.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 64 | <=2025-11-25 | v2 |
Return URL clips, deduped by host.
List recent clipboard entries, newest first. Secret values are never returned — only metadata.
Pin or unpin a clipboard entry.
Full-text search clipboard history (FTS5 BM25).
Add or remove a tag from a clipboard entry.
Decrypt and emit a stored secret. Requires biometric authentication (Touch ID on macOS).
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Despite clear risk stratification (READ_ONLY, WRITE, DESTRUCTIVE), annotations are not declared in tool definitions.
clear_history scope parameter accepts free-form strings ('all', 'older-than-days:N', 'kind:K') without enum constraint. LLMs may hallucinate invalid scope values.