Reported issues for netlogger-mcp

Pod holds 8 of 8 GitHub reports that passed its relevance review. This can include external user reports, maintainer-confirmed bugs, and concrete feature gaps. Treat them as evidence to inspect, not a count of distinct defects.

Back to netlogger-mcp.

Most discussed

Normalized band and frequency beside NetLogger's raw values

Found in Patton's cold-seat test (2026-09-28). NetLogger's fields are free text entered by loggers:

• band: 40m, 40M, 75M
• frequency: 7033 (kHz), 3918 KHz, 7.272 (MHz), 147.030 91.5 (with a CTCSS tone), TG 3141 (a DMR talkgroup)
• net_control: sometimes missing, and sometimes not a callsign (ROB)
• current_name: cut at 32 characters by NetLogger

Proposal: keep the raw values, and add fields where they can be read reliably. Leave a field out when the text…

Read the thread · 2026-09-28 · closed · outside contributor · 1 comment

Keep to NetLogger's call limits

NetLogger's published guidance (spec v1.2+), which this server must never exceed:

• A per-call limit over a rolling 60 seconds, enforced before any request is sent.
• Cache every answer: active nets 60 s (filter locally, so any number of name filters cost one call), check-ins 20 s per net, past nets 60 s per query, past check-ins an hour (they don't change).
• Over the…

Read the thread · 2026-09-28 · closed · external user · 1 comment

Contract: one net and check-in record, whatever the source

Tools return records defined here, never a source's raw format, so the same answers come from NetLogger today and from the OM-Logger OMISS is migrating to.

• Net: source, server, name, current name (NetLogger's AltNetName), frequency, band, mode, net control, logger, opened, monitoring count; for past nets also the net ID, closed, last activity, AIM on/off, auto-closed.
• Check-in: serial number, callsign, name and preferred name, city, county, state, country, DXCC, grid, status,…

Read the thread · 2026-09-28 · closed · external user · 1 comment

Check-ins: note rows are not pending rows

From tonight's OMISS 40m net (2026-09-29, net control KC9RRN), compared with the net's published report.

Rows 78–85 have no callsign, and netlogger-mcp 0.1.4 lists them all in pending_serials. But most of them carry the logger's own notes, in the member-ID and remarks fields: # # NET START: 01:00, # # FREQUENCY: 7.192, # # NET CLOSE: 02:41, # # NO LAST CALLS TONIGHT BAND DROPPING OUT. Only the rows with nothing at all are really pending (or blank spacers).

Fix:

• **no callsign,…

Read the thread · 2026-09-29 · open · outside contributor · 0 comments

Check-ins: drop empty slots from the list and count; name the station at the pointer

Found by Patton on a live net (OMISS 40m, 2026-09-29 00:37Z, NETLOGGER2):

1. Empty slots are passed through as check-ins. Serials 21 and 25 came back as {"source", "serial"} with no callsign, and checkin_count (20) counted them. A record with no callsign isn't a check-in: leave it out of checkins, count only real ones, and report empty_slots (their serials).
2. Serials move between calls (W2EUA went from 25 to 24 as the logger edited the list). The pointer is a serial, so the…

Read the thread · 2026-09-29 · closed · outside contributor · 0 comments

Share cached answers between copies, not just the call budget; as_of_utc on every response

From Patton's cold test of omiss-mcp 0.1.1 (2026-09-28). The call budget is shared by every copy through the limits file, but the answer cache is per process. So when one server (netlogger-mcp) has just fetched active nets, another (omiss-mcp) that hits the shared limit returns an error, even though a 49-second-old answer exists on the same computer. The shared budget was meant to give users cached data instead of errors.

Fix:

• Shared answer cache: a small file beside limits.json,…

Read the thread · 2026-09-28 · closed · outside contributor · 0 comments

Samples are synthetic: transcribe fixtures from real NetLogger responses

The four sample XML files in src/netlogger_mcp/samples/ are synthetic — written to the shape the spec describes, not transcribed from responses NetLogger actually sends. Every parser test therefore checks the code against our own reading of the spec, which is the one thing it cannot falsify.

Why this is worth doing now rather than later

qrz-mcp shipped three bugs of exactly this kind, all found by outside contributors (@MicaelJarniac, @ssamjung2) against a live logbook, and all…

Read the thread · 2026-09-28 · open · 0 comments

Share the call limits between copies of the server on one machine

The limits are per process today. Two AI apps on one computer (e.g. Claude Desktop and Claude Code) each run their own copy, and each keeps to NetLogger's limits on its own, so together they could make twice the calls. Every request carries the user's callsign, so NetLogger sees both as one station.

Share the budget across processes on one machine: a small lock-protected state file in the settings folder, or one local copy that the others talk to. It has to work on Windows, macOS and Linux.…

Read the thread · 2026-09-28 · closed · external user · 0 comments

Most recent

The remaining reports are on the project's issue tracker.