io.github.sbrunner-atx/wsjtx-mcp
An MCP server that controls WSJT-X for amateur radio weak-signal modes (FT8/FT4, etc.) via UDP, enabling observation of status and decodes, and transmit operations like replying to CQs or sending free text.
README
<!-- mcp-name: io.github.sbrunner-atx/wsjtx-mcp -->
wsjtx-mcp
An MCP server that controls WSJT-X (FT8/FT4/JT65/MSK144/Q65/WSPR…) from MCP clients such as Claude Desktop and the MCP Inspector.
It is the weak-signal leg of an "operate → log" trio for amateur radio:
- fldigi-mcp — operate broad digital modes via fldigi (XML-RPC).
- contest-mcp — log QSOs to N3FJP (TCP API).
- wsjtx-mcp (this one) — operate the FT8/FT4 weak-signal world via WSJT-X (UDP message protocol).
⚠️ Experimental (v0.1). Transmit is gated behind your callsign and you keep the operator in command. Read the Transmit safety section.
How it is different
WSJT-X does not offer a request/response API. It broadcasts state over UDP
(Status, Decode, QSOLogged, Heartbeat, …) and honours a small set of
inbound control messages. So this server runs a background UDP listener that
continuously parses datagrams and keeps the latest status, a buffer of decodes,
and completed QSOs — you read those, and nudge WSJT-X with control messages.
Consequences worth knowing up front:
- No dial-frequency control over UDP. You can read the dial frequency from
Status, and set mode/sub-mode/Rx DF/T-R period viaconfigure, but QSY is a rig-control concern (Hamlib/CAT or the UI), not this server. - You can start and halt Tx, but cannot toggle "Enable Tx". Transmission is
started by answering a CQ (
reply) or byfree_textwithsend=true, and stopped bytransmit halt. There is no UDP command for the "Enable Tx", "Auto Seq", "Call 1st", or "Hold Tx Freq" checkboxes — those stay UI settings. replygives hands-free QSOs only when WSJT-X's "Auto Seq" is on. Areplyis equivalent to double-clicking a CQ; with Auto Seq enabled (the usual FT8/FT4 default) WSJT-X then sequences the whole exchange toQSOLoggedwith no further calls. With Auto Seq off,replystarts only the first transmission.- Best for search-and-pounce, weak for RUN. The API is built to answer CQs
(and only CQ/QRZ decodes), so S&P is fully automatable. It has no clean way to
drive a repeating call-CQ (RUN) cycle — that needs WSJT-X's own "Enable Tx",
or a
free_textCQ re-sent each period.
Requirements
- WSJT-X 2.1 through 3.x — the UDP message protocol is schema 3 / Qt_5_4 and has been stable across those releases (verified live against WSJT-X 3.0.2).
- In WSJT-X: Settings → Reporting → UDP Server
- UDP Server = the host running this server (default
127.0.0.1), port2237. - Accept UDP requests = ON to allow control (it is OFF by default). Observing decodes/status works without it; commanding does not.
- UDP Server = the host running this server (default
Install (Claude Desktop)
Download the wsjtx-mcp.mcpb from the
latest release and
double-click it, or drag it onto Claude Desktop → Settings → Extensions. Fill in
the settings form (callsign, host, port). See docs/INSTALL.md.
Configuration
| Variable | Default | Purpose |
|---|---|---|
WSJTX_HOST |
127.0.0.1 |
UDP address to bind/listen on. |
WSJTX_PORT |
2237 |
WSJT-X UDP Server port. |
WSJTX_CALLSIGN |
(empty) | Operator callsign — the single transmit gate. Blank = receive-only. |
WSJTX_MULTICAST |
(off) | Optional multicast group to join (coexist with other UDP consumers). |
WSJTX_INSTANCE |
(auto) | Target a specific WSJT-X Id when several instances broadcast. |
Host/port are where this server listens; control replies are sent back to the address each datagram arrived from.
Tools
| Tool | Kind | What it does |
|---|---|---|
status |
observe | Latest Status snapshot + listener/instance health. |
diagnostics |
observe | Host/network + bind status + datagram counts + gate state. |
decodes |
observe/nudge | read / drain (poll new) / clear_local / replay. The RX plane. |
log |
observe | Buffered completed QSOs (QSOLogged + LoggedADIF) → feed N3FJP. |
reply |
transmit | Answer a buffered CQ/QRZ decode (auto-sequences the QSO when WSJT-X "Auto Seq" is on). |
free_text |
transmit if send |
Set the Tx5 free-text message; send=true keys the radio. |
transmit |
control | halt / halt_auto — stop transmitting (UDP can't enable Tx). |
configure |
control | Mode/sub-mode/Rx DF/T-R period/freq-tol/DX call+grid. No dial freq. |
clear |
control | Clear the Band Activity / Rx Frequency windows. |
highlight |
control | Colour or clear a callsign in Band Activity. |
location |
control | Override the session Maidenhead grid. |
switch_config |
control | Switch to a named WSJT-X configuration. |
annotate |
control | Set a Fox/Hound sort-order annotation for a DX call (niche, DXpedition). |
wsjtx_call |
escape hatch | Build & send any message type by name (gate still applies). |
Transmit safety
The callsign is the single transmit gate, exactly as in fldigi-mcp. With
WSJTX_CALLSIGN blank the server is receive-only: it refuses every
transmit-initiating message — reply, free_text with send=true, and any
keying message via wsjtx_call. transmit halt, clear, configure,
highlight, location, replay, and all reads are always available (they don't
put you on the air; halt takes you off).
Beyond that gate:
- Per-transmit approval comes from the Claude Desktop tool-permission prompt — lean on it for human-in-the-loop control.
- WSJT-X's own Tx Watchdog and the
Tx Enabled/Transmittingflags (surfaced instatus) are extra safety signals. - Operating under Part 97 automatic/remote control is the operator's responsibility: ensure station identification and a control operator who can intervene.
Running alongside other UDP tools
Only one process can normally own UDP 2237 on a host. If JTAlert, GridTracker,
or N1MM already consume it, either point WSJT-X's secondary UDP server here, use
a multicast group (WSJTX_MULTICAST) so several listeners coexist, or run
this server on a different host. To reach a WSJT-X on another machine, install
the mcp-host-bridge tool
(mcp-host-bridge install wsjtx --to <rig-host>) and set WSJTX_PORT=2238 —
sandboxed MCP clients reach only loopback, and the bridge does the LAN hop. See
docs/REMOTE-HOST.md.
Development
uv sync
uv run ruff check .
uv run pytest
The protocol codec is pure standard library and unit-tested against byte
fixtures, so the tests need no running WSJT-X. A smoke_test.py proves a live
WSJT-X is reachable receive-only. The field-tested message reference lives in
docs/WSJTX-API.md.
License
MIT © 2026 Stefan Brunner (AE5VG)
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.