Loading ToolBench

Reading the current corpus…

mom MCP server · F · ToolBench

mom

Memory Oriented Machine — persistent memory for AI agents. A Go-based CLI tool and daemon that manages operational event ingestion, session tracking, and memory projection for AI agent systems.

2026-07-28+STREAMABLE_HTTP7 tools

27/100

Definition Quality23

Protocol Readiness0

Supportability50

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence · evidence: Streamable HTTP

Current-spec patterns detected

Summary

This HTTP server exposes 7 tools via REST endpoints, but lacks critical MCP-compliant tool definitions. No explicit tool registration with schemas is visible in the provided source. Tool names are HTTP routes (POST /api/ingest/events), not verb_noun patterns. Descriptions exist but are generic and lack LLM-optimization guidance. Input parameters are documented in descriptions but lack formal JSON Schema with type constraints. Output schemas are not documented. Error handling is absent from the code sample. The server is a raw HTTP API, not an MCP server, it does not implement the MCP protocol at all.

Tools (7)

GET /api/ingest/eventsread onlysource verified32/100

Iterate and filter ledger records. Supports filtering by project_id, type_prefix, and offset range. Returns paginated results with optional next offset for continuation.

GET /api/ingest/headread onlysource verified20/100

Get the current head offset of the ledger.

GET /api/ingest/healthread onlysource verified18/100

Health check endpoint that returns server status and version information.

GET /api/metaread onlysource verified15/100

Get metadata about the lens dashboard including total sessions and total memories.

GET /api/sessionsread onlysource verified18/100

List all sessions with summary information including interactions, memory count, tool usage, tokens, and cost. Results are sorted by start time (newest first).

GET /api/sessions/{id}read onlysource verified27/100

Get detailed information for a specific session including tool calls breakdown and memory items.

POST /api/ingest/eventswrite

Issues (5)

CRITICALtool ↗

Not an MCP server. This is a raw HTTP API with no MCP protocol implementation. No tool registration, no capabilities negotiation, no MCP message format.

POST /api/ingest/eventsGET /api/ingest/eventsGET /api/ingest/headGET /api/ingest/healthGET /api/metaGET /api/sessionsGET /api/sessions/{id}

HIGHtool ↗

Tool names are HTTP routes, not verb_noun patterns. LLMs cannot infer intent from 'POST /api/ingest/events', should be 'append_operational_events' or 'ingest_events'.

POST /api/ingest/eventsGET /api/ingest/eventsGET /api/ingest/headGET /api/ingest/healthGET /api/metaGET /api/sessionsGET /api/sessions/{id}

HIGHconstrained-input ↗

No formal JSON Schema for inputs. Parameters documented in prose descriptions only. GET /api/ingest/head and GET /api/ingest/health have empty input objects with no schema definition.

GET /api/ingest/head

Recommendations

Wrap this HTTP API in an MCP server layer. Implement the MCP protocol (stateless request/response, _meta with logLevel, capabilities negotiation) on top of the existing HTTP endpoints.
Rename tools to verb_noun patterns: 'append_operational_events', 'list_operational_events', 'get_ledger_head', 'check_ingest_health', 'get_lens_metadata', 'list_sessions', 'get_session_details'.
Define formal JSON Schema for all tool inputs. For POST /api/ingest/events, expose the ingestItem struct as a schema with type='object', properties={type, payload}, required=['type', 'payload']. Add enum constraint for type prefix.
Document output schemas. Expose postResponse, getResponse, headResponse, healthResponse as JSON Schema so LLMs know what fields to expect and can chain tools.
Add error handling with recovery hints. Return 400 with 'Invalid event type: must start with "operational." prefix' instead of a generic error. Return 503 with 'Ledger unavailable; retry in 5 seconds' for service failures.

Score history

Overall score trend

First recorded score · v2 rubric

27/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	27	2026-07-28+	v2