gledati (slav.): to watch.
what a good analyst does before speaking.
A shelf of free tools. Each one watches something public and hands back an answer.
Same author, different shapes. They share no stack, no interface and no release cadence. What they share is the promise: free, MIT licensed, installable, running on your own machine, no account to start, and a written audit of what each one cannot see.
Five of twelve slots are filled.
| Tool | Watches | Tells you | |
|---|---|---|---|
| 01 | signal | Thirteen public sources around whichever market you name | What changed, scored and attributed, over MCP |
| 02 | signal-patterns | Nothing. It is the written reasoning | What is watchable, what it costs, how to build it |
| 03 | whenrun | Day-ahead power prices and renewable share | The cheapest or greenest window, then starts the job in it |
| 04 | phonehome | Every DNS query on your own network | Which device talked where, and when that changed |
| 05 | recall | Four public registers, against a list of what you own | Which of your things has a notice, and how sure the match is |
| 06 | [UNNAMED] |
[UNDECIDED] |
[UNDECIDED] |
Slot 06 is empty and stays written that way until something runs there. Is there something you keep checking by hand that a tool could watch for you? Ask for it in the open.
Track anyone. Compare anyone. On your own machine.
Anyone means any company you name in one file. signal watches what they publish, scores and attributes it, correlates it into convergences that several independent publishers corroborate, and answers your agent over MCP rather than handing a person one more dashboard to read. The dashboard is one client, not the product.
┌──────────────────────────┐ ┌────────────┐ ┌──────────────────────┐
│ ingest │──▶│ turso │──▶│ battlecards │
│ news · HN · reddit │ │ (libSQL) │ │ analyst briefs │
│ youtube · github │ └─────┬──────┘ │ convergence alerts │
│ certs · sitemaps │ │ └──────────────────────┘
│ trends · search · email │ ▼
└──────────────────────────┘ ┌───────────┐ ┌──────────────────────┐
│ analyst │ │ MCP surface │
│ /scan │ │ 8 tools, read only │
│ /deep │ │ until you say │
│ /gap │ │ otherwise │
└───────────┘ └──────────────────────┘
The unit that matters is a convergence, not a mention. One correlated pattern that several independent publishers corroborate, scored from its own evidence, with the rule that fired it stated inline. Ten raw mentions of the same press release are one event.
13 AI coding agents and app builders tracked, across two segments
13 signal sources, from RSS to certificate transparency logs
2,600+ signals in the deployment behind the live preview
8 MCP tools, read only until you say otherwise
3 runtime dependencies, no build step, clone and run offline
Clone it, run npm run db:migrate, and you have a populated dashboard with no account and
no API key. Point it at your own companies by editing one gitignored file.
runtime · Node 22, zero dependency bias. three runtime deps, no build step
storage · libSQL. a local file by default, hosted only when you want it shared
ai · bring your own key. without one, a keyword classifier and no briefs
agents · MCP over stdio. read only by default, spend ceiling shared with cron
tests · offline, against fixtures captured from the real sources
local first. the laptop is the runtime. cloud is for sharing state between machines you own.
convergence over alerts. one correlated insight beats ten raw mentions.
say what you cannot see. every tool ships an honest audit of its own blind spots.
read only until asked. an agent does not spend your money because it found a button.
generated text is labelled. model output is marked as model output.
idempotent retry, not dual write. retry until it sticks. never split the source of truth.
__ __ __
____ _ / /___ ____/ /___ ______/ /_
/ __ `// // _ \/ __ // __ `/ ___/ __ \
/ /_/ // // __// /_/ // /_/ / /__/ / / /
\__, //_/ \___/ \__,_/ \__,_/\___/_/ /_/
/____/
if it ships to turso, it shipped.