DRM3 Enterprise · Signalpoint
Signals out of the record.
- Address
- signalpoint.io
- Sign-in
- A DRM3 account. Seats are assigned.
- The door
- The form on this page. A person reads every note.
Who it is for
Act on it, or sell it
For teams that act on public records
Know what moved, before the press does.
Government affairs, family offices, construction, finance, procurement: a desk that reads the records of the jurisdictions you follow and files what matters, scored and staged, with the source behind it. Your people search it by meaning, keep watchlists, and get what moved by email or into their own tools.
What a signal isFor holders who sell access
Sell the signals. Readers subscribe.
A desk on your brand where readers search your record by meaning, keep watchlists and get a daily or weekly digest. The same record as an API for their own tools. You set the plans and the prices: one checkout, one monthly statement.
How we set it upHow it works
Four steps, and we run three of them
- 01
Bring it
Send the records the way they already move: an API push, a storage bucket, a file drop or SFTP.
- 02
We read and score it
Every record is read for the matter in it, scored 0 to 100 with a decision stage, filed by subject and indexed by meaning. A record that carries no matter is filed as nothing.
- 03
Build on it
The signals become watchlists, briefs, research across the whole set, and an API and an agent door your own tools can read.
- 04
People act on it
Your team, or your readers on your brand, see what moved and open the source record behind it.
Signals
A signal is the matter in a record.
A meeting record runs for hours. The matter in it is a contract approved, a rezoning passed, a bond funded, a board seat filled. Signalpoint files that, and nothing else.
What a signal carries
A headline and a summary
Written about the matter, not the meeting: what was decided, for whom, for how much.
A score, 0 to 100
How material the matter is to the subject it was filed under. A desk's wall runs most material first.
A decision stage
Early, discussed, decided, funded. The stage says how late you are.
Where and when
The jurisdiction, the body, the meeting and its date, so a watchlist by place is exact.
The source
A link into the record at the place the matter was raised. A reader opens it and compares.
What a signal is not
A proclamation, an award, a public comment with no decision behind it. The reading files nothing for those, so a watchlist is quiet until something moves.
A guess. A signal that cannot be tied to a place in the record is not filed.
Why the search finds it
The record is indexed by meaning. A reader who searches for a variance finds the rezoning filed under land use; a reader who searches for ambulance funding finds the budget line that carried it. The words typed are a starting point, not the boundary.
Every answer the desk gives, a brief, a summary, an answer to a question, cites the record it came from.
How it works
We set it up for you.
You bring the records, yours or a customer's. We take them in, read and score them, stand up a desk on the right brand and address, and hand over the keys. A desk is live in days.
The setup, step by step
- 01
You bring the data.
Yours, or a customer's. Tell us what it is, how it moves, and who should see it: buyers who will subscribe, or your own team that wants alerts. One conversation covers it.
- 02
We take the data in.
The export arrives the way it already moves: an API push, a storage bucket, a scheduled file drop or SFTP. Your engineer's part is about a week.
- 03
We refine it.
Every record is graded and scored, then indexed so a buyer can search it by meaning. Each record becomes a signal with its source behind it. A correction replaces the version before it, and the earlier version stays on the record.
- 04
We put it on the right brand.
Your name or your customer's: the mark, the colours, the address. One quiet line at the foot is all that says DRM3.
- 05
Buyers sign in and subscribe.
They search the record, keep watchlists, and get what is new by email, daily or weekly.
- 06
You get a statement every month.
The customer pays through one checkout, and you earn a share of every plan.
What your customer gets
Search by meaning
Search the record for what you mean and narrow it by section, jurisdiction and date. Open any signal to its source.
Signals, watchlists, digests
Scored signals filed by subject, watchlists that mail what moved, a bell in the reader's account, and an ask-the-desk answer that cites the record.
An API with a playground
The same record as data, for the customer's own tools and for agents.
Their own admin console
Sections, readers, invites and settings, in English or Spanish.
The three questions everyone asks
Is there a setup fee?
No. There is no build fee either. A customer's desk is a plan with caps, priced once, and it does not move under that customer.
What does it cost to bring my data in?
Nothing. Your export lands through a keyed door that already exists. Your engineer's week is the whole of it.
What comes off the top?
One share, on a monthly statement, set once before the first deal and written in. It covers support, operations, onboarding and customization. You keep the larger part.
How data comes in
Send it the way it already moves. We refine it from there.
Your team does not need a new pipeline to open a new channel. Tell us how the data leaves your systems today. We wire that path in, and the refinery takes over.
Ways to send data in
API push
Post rows to a keyed door over HTTPS. Best for data that changes through the day.
Storage bucket
Drop exports in a bucket the desk reads on a schedule. Best for nightly or hourly batches.
File drops
Send CSV or JSON files on a schedule: a full dump, or only what changed.
SFTP
Push files to an SFTP drop, if that is how your systems already deliver.
Upload by hand
Upload a CSV, or a zip of CSVs, from the Data tab of your desk. Good for a first look.
Have a path that is not on this list? Tell us. If your systems can export it, we can take it in.
What the refinery does
- 01
Grade.
Each record is read for what it is: a meeting, a filing, a transaction, a notice. The fields that say what happened, when and where are found and kept.
- 02
Score.
Each record is weighed against the rest. Weak or duplicate rows fall away. Strong rows stand, each on its own source.
- 03
Make it searchable.
The record is indexed by meaning, so a buyer finds what they mean and not only the words they typed.
- 04
Make it buildable.
The refined record becomes signals filed by subject, briefs for a watchlist, research across the whole set, and an API that other tools read.
What we need from you
A sample of the data, or a description of its fields.
Which fields say what a record is, when it happened and where it came from.
How often it changes.
Who may read it.
Updates and corrections
Each delivery is refined and indexed the same way. Send a full replacement or only the changes. A correction replaces the version before it, and the earlier version stays on the record, so a buyer can always see what changed.
How buyers read it
The desk
Search by meaning, filters, and a page for every signal with its source behind it.
Alerts and digests
Watchlists that email a buyer when something new matches, daily or weekly.
The API
Signals, search and sections as JSON, with a playground and an OpenAPI document.
Agents
An MCP door for AI agents, and a plain-text view of the desk a model can read.
The key is yours
The desk reads your data with a key that belongs to that desk alone, and you can revoke it at any time.
Security
Your data stays yours.
Data you sell carries more risk than most. So the protection is part of how a desk works, and you can check each part.
One key per desk
A desk reads its dataset through the same metered door any outside app would use, with a key that belongs to that desk alone. Signalpoint holds no backstage key to your data. Revoke the key and the desk stops reading.
Private in the query
A customer's records are marked private where they are stored. Every public page asks for records that are not private, in the query itself. A setting flipped by mistake cannot put them on a public page.
Behind a sign-in
A desk sits behind a sign-in wall. Buyers are invited by email, and the roster is the customer's to manage. An address that is removed meets the wall on its next request.
Two-step sign-in for admins
A desk can require that its admins sign in with a passkey or an authenticator code. Admins are the people who invite buyers, export the audience and hold the API keys. With the rule on, a password alone opens nothing on that desk for an admin. Only an admin who already signed in that way can change the rule, and no API key can.
Never mixed with anyone else's data
A customer's records do not enter shared search, another desk, or any data pack. One desk reads one customer's data.
Yours to take back
Delete a dataset and every door to it shuts at once: the keys, the shares and the API. The data itself is removed right away, or after a one-day grace period if you want a way back.
The source on every signal
Each signal shows the source record it stands on, so a buyer can open it and compare.
Questions from your security team? Send them through the form below.
Sources
The source behind every signal.
A buyer should not have to take a signal on faith. Each one shows the record it came from, and each desk shows where its data comes from.
What a buyer sees on a signal
Every signal has its own page. The page shows the source record the signal stands on, and a link into the exact place in that record. The reader can open the source and compare the two.
Where the data comes from
Every desk has a Data and provenance page. It lists each feed the desk stands on: who supplies it, how it arrives, how many records it holds and when it last delivered. A buyer can see the whole supply line before relying on a signal.
Corrections stay visible
When a record is corrected, the new version replaces the old one on the desk, and the earlier version stays on the record. Nothing is overwritten.
Private stays private
Showing a source to a signed-in buyer does not publish it. A customer's records stay behind the desk's sign-in.
API and agents
Run a desk with an agent.
Everything the console does, the API does too. An agent with an admin key can point a desk at its data, build the signals, invite readers and send the digest.
Get a key
Open your desk and mint an admin key under Desk settings. The key is shown once, and it works on its own desk. No desk yet? Use the form below.
Connect over MCP
Point any MCP client at your desk's MCP door and pass the key. In Claude Code:
claude mcp add --transport http signalpoint https://<your-desk>/mcp \
--header "Authorization: Bearer YOUR_ADMIN_KEY"The descriptor is at /.well-known/mcp.json on your desk's address. Every tool is also a JSON route, called with the same key in an Authorization: Bearer header.
What an agent can do
set_signal_desk and get_signal_desk point a desk at its data and read the setup back.
build_signals reads the data and files one signal per source row. signal_build_status shows each build.
invite_reader, remove_reader and signal_access manage who can read the desk.
ask_desk answers a question from the desk's own record. summarize_signal sums up one signal, with citations.
brief_watchlist and send_digest tell readers what is new.
The rules the API keeps
A private desk answers a request with no key with a refusal, never with data.
A key reads and writes its own desk.
A customer's records never leave the desk they belong to.
A guided build
The Signalpoint skill walks an agent from a fresh key to a running desk. It is served from your desk's address at /agents/skill.md.
Start a conversation
What do you need to know, and from what records?
Say which records you follow or hold, who should read the signals, and what a signal has to tell them. A person reads every note and answers with what the desk would look like.
The sign-in for every DRM3 solution is a DRM3 account. We seat the first admins once the agreement is in place.
Signalpoint is one of the solutions under DRM3 Enterprise. It runs on DRM3 Rails, the white-label pattern Newsroom Floor also uses.