cablegram-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CABLEGRAM_DB | No | Override the default archive database location. Defaults to ~/.local/share/cablegram/archive.db on Linux. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| wire_latestA | What 19 tech/AI sources published in a time window, grouped by source, newest first within each. English, Chinese and Russian, untranslated. Filtered by DATE ONLY — never ranked, never scored. The order says nothing about importance. A source listed as DOWN means UNKNOWN, not 'nothing happened there'. A declared CUT (hn=25/57) means more exist in the window; raise limit_per_source or narrow the window. CROSS counts how many sources carried the same URL. It is arithmetic, not a ranking — but a story in six feeds across three languages within hours is the earliest signal this server can give you. detail='headlines' (the default) returns titles and ids; pass those ids to wire_read for the text. detail='full' includes each stored body inline and drops to 5 per source, because bodies are expensive — a teaser is marked as one, and is NOT the article. |
| wire_readA | The stored text of specific dispatches, by the ids wire_latest or wire_search returned. Read body=teaser literally: that feed ships a truncated excerpt, and the text you get is NOT the article. Do not draw conclusions from it — open the url or say the full text was not available. Ids not in the archive are named in the reply rather than dropped: re-run wire_latest or wire_search for the same window to get current ones. |
| wire_searchA | Search the archived headlines of every source that carried a story. IMPORTANT: this searches only what this server has archived since it was first run — not the whole internet and not the sources' own history. '0 hits' means 'not in what we can search'. It does NOT mean nobody is talking about it, and must never be reported as such. Chinese and Russian sources are indexed in their own language: a company is 智谱 here and Zhipu on Hacker News. If a query comes back empty, retry it transliterated or translated before concluding anything. |
| wire_sourcesA | The catalogue and its health: which sources exist, their language and tags, when each last answered, and which are failing. Read this before concluding a topic is quiet. A source that has never been polled holds nothing, which is a different fact from a source that holds nothing. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 4 tools
Each tool has a clearly distinct role: browsing by time window, searching archived headlines, reading stored dispatches, and checking source health. The overlap between wire_latest and wire_search is resolved by their descriptions (chronological feed vs archive query).
All names share the wire_ prefix and use lowercase snake_case, so the set feels predictable and cohesive. The second element is not consistently a verb or noun (latest, read, search, sources), but the pattern is still clear enough to be navigated without confusion.
Four tools is well-scoped for a read-only news/archive monitoring server. Each tool earns its place and there is no bloat.
The server covers its apparent workflow completely: discover via latest or search, retrieve via read, and assess archive coverage/health via sources. The descriptions also explicitly address edge cases like teasers and unindexed queries, leaving no dead ends in normal use.