Model Context Protocol server for GoLogX audit logs. Exposes three tools that wrap the GoLogX audit package to verify hash-chained logs, append tamper-evident entries, and read entries back after verification.
logx-mcp presents three audit-focused tools with clear, domain-specific names and well-documented schemas. All tools follow verb_noun naming (verify_, append_, read_) and include comprehensive input schemas with typed parameters and descriptions. However, output schemas are documented only implicitly through code comments rather than formal schema declarations. Error handling is present but generic in places. The tool set is coherent and narrowly focused on cryptographic log verification, which is a strength for composition but limits applicability.
Append one tamper-evident entry to a GoLogX audit log, creating the file if needed and resuming the existing hash chain across calls. Optionally signs the entry with an Ed25519 private key PEM. Each call adds exactly one hash-chained record.
Read entries back from a GoLogX audit log. The chain is verified first, so tampered data is never surfaced as trustworthy; the result reports the verification outcome alongside the entries. Use limit to return only the most recent entries.
Verify the integrity of a GoLogX hash-chained audit log file. Reports whether the chain is intact or, if not, the index and detail of the first entry that was edited, deleted, reordered, or forged. Optionally checks Ed25519 signatures when a public key PEM path is given.
Output schemas not formally documented. Tool descriptions reference returned JSON shapes (e.g., 'ok', 'entries', 'signed', 'mode' fields) only in prose comments, not in explicit output schema definitions. LLMs cannot parse unstructured descriptions to reliably extract fields for chaining.
Error messages lack recovery guidance. When a file is unreadable or a key is invalid, errors return raw messages like 'could not read /path: permission denied' without suggesting next steps (e.g., 'Verify file permissions' or 'Check public key path format').
No pagination or limit enforcement on read_audit_log. The 'limit' parameter accepts 0 for 'all entries' with no upper bound documented. A malicious or confused agent could request millions of entries, consuming unbounded memory and tokens.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | 2025-06-18+ | v2 |
Tool descriptions use technical cryptographic language ('hash-chained', 'tamper-evident', 'Ed25519') but lack usage guidance for LLMs unfamiliar with audit log patterns. No description explains WHEN to call verify vs. read, or what state assertions each provides.