Developers

An API and an MCP server are planned.

Q: What would the API cover?

Reading the wire, a record's trust and bias readings and six-axis audit, dossier export, webhooks for flashes and tier changes, and exports in JSON, CSV and PDF. Single sign-on with SAML or OIDC is a separate planned item.

Q: Will the API return scores without their inputs?

No. The design principle is that provenance travels with every record: a score comes with its inputs and a claim comes with the count of origins behind it.

A self-serve API and an MCP server, a connector that lets AI agents ask for records, are planned and not built. Nothing here can be called yet.

PlannedNot availableBook a call to talk through the API and MCP server with us.

Book a call Tell us what you would build What exists today

DetailsWhat this page describes

Planned, not built

This page describes a self-serve API and an MCP server for agents and developers, so that you can tell us what would matter before they are designed. Neither is built. Nothing here can be called.

Nothing on this page can be called yet. No endpoint, key, sandbox, beta or MCP server is open. The page says what we plan to build so that you can tell us what would matter to you before it is designed.

Interested? Fill in the form below.

02 A sketch

What a call could look like.

Pick a sketch to see a request and its answer. Shape only: the hosts and paths are placeholders and none of it is live.

A list of dispatches, each with the provenance counts that sit behind it.

Request

GET <api-host>/v0/wire?zone=Z06&since=2026-09-14T00:00:00Z&limit=2
Authorization: Bearer <token, not issued>

Response

{
  "items": [
    {
      "record_id": "7A-D-000731",
      "headline": "Harbour authority closes main container port, shippers say",
      "filed_at": "2026-09-14T03:50:00Z",
      "tier": 1,
      "provenance": { "reports": 14, "origins": 3, "primary_documents": 1 },
      "trust": { "score": 74, "see": "/v0/records/7A-D-000731/audit" }
    }
  ],
  "next": "<opaque cursor>"
}

Sketch, not a live endpoint

DetailsAbout these sketches

Four illustrations of shape, not of a contract. The hosts, tokens and signature scheme are placeholders, the paths are not final, and the record numbers are examples. Do not build against them. The full sketches are listed in the page text further down.

Read the wire
Audits
Dossier export
An MCP server for agents
Webhooks and alerts
Single sign-on
Data export formats

03 Coverage

What it could cover.

These are the areas we expect to matter, all planned. Which come first depends on what you tell us.

Read the full explanation

01Built today, planned next

What exists, and what does not.

The full list is on the roadmap. Access and pricing are on the pricing page.

Built
The wire, the record, the audit and the dossier, as screens
Working today. You can use them by hand.
Live demo
Built
A live data service behind the wire
Live data sources and the full backend are in place. An API would sit on this service.
Roadmap
Planned
A self-serve API
Described on this page. Not designed in detail and not built.
Roadmap
Planned
An MCP server for agents
Not designed in detail and not built. It would sit on the same service as the API.
Roadmap
Planned
Webhooks and alerts
Not built. It would sit on the live data service.
Roadmap
Planned
Single sign-on, SAML or OIDC
Not built. Listed on the roadmap as a planned item.
Pricing and access
Planned
Programmatic access terms and pricing
Not decided. No price is published for anything, including this.
Pricing and access

02What the API and MCP server could cover

Planned

What it could cover.

These are the areas we expect to matter. Which come first depends on what you tell us.

Read the wire: List by , time and , newest first, with the of each record. Read-only, with paging. It would cover the sectors that exist when it ships: Geopolitics & Borders today, with 15 sectors planned.
Audits: Fetch one record's trust and bias readings with their inputs, and its with the confidence and the contested mark for each axis.
Dossier export: Ask for a record as a in a stated format, and read back whether it has been and when.
An MCP server for agents: A Model Context Protocol server, so an AI agent or a developer tool can ask 7Analytica for records and audits directly, with the same a person sees. Planned, not built, and no tools or commands are defined yet.
Webhooks and alerts: Be told when a flash is raised or a changes, with the rule that raised it, instead of polling.
Single sign-on: SAML or OIDC sign-in for organisations that need to connect their own identity provider. This is also on the roadmap as a planned item.
Data export formats: JSON for records and audits, CSV for tables of readings, and the PDF 7Analytica already makes.

03Sketches

What a call could look like.

Four illustrations of shape, not of a contract. The hosts, tokens and signature scheme are placeholders, the paths are not final, and the record numbers are examples. Do not build against them.

01Read the wire

Sketch, not a live endpoint

A list of dispatches, each with the provenance counts that sit behind it.

Request

GET <api-host>/v0/wire?zone=Z06&since=2026-09-14T00:00:00Z&limit=2
Authorization: Bearer <token, not issued>

Response

{
  "items": [
    {
      "record_id": "7A-D-000731",
      "headline": "Harbour authority closes main container port, shippers say",
      "filed_at": "2026-09-14T03:50:00Z",
      "tier": 1,
      "provenance": { "reports": 14, "origins": 3, "primary_documents": 1 },
      "trust": { "score": 74, "see": "/v0/records/7A-D-000731/audit" }
    }
  ],
  "next": "<opaque cursor>"
}

02Read an audit

Sketch, not a live endpoint

The six axes for one record. A contested axis is marked, never averaged away.

Request

GET <api-host>/v0/records/7A-D-000731/audit
Authorization: Bearer <token, not issued>

Response

{
  "record_id": "7A-D-000731",
  "axes": [
    { "axis": "who",   "confidence": 86, "contested": false },
    { "axis": "where", "confidence": 58, "contested": true  },
    { "axis": "why",   "confidence": 38, "contested": true  }
  ],
  "trust": {
    "score": 74,
    "inputs": { "longevity": 80, "corroboration": 70, "ownership": 70, "veracity": 80 },
    "formula": "round((longevity + corroboration + ownership + veracity) / 4)"
  }
}

03Export a dossier

Sketch, not a live endpoint

Asks for a file in a stated format. A sealed dossier is never rewritten, so a later change is a new entry.

Request

POST <api-host>/v0/records/7A-D-000731/dossiers
Authorization: Bearer <token, not issued>
Content-Type: application/json

{ "format": "pdf", "theme": "light", "seal": true }

Response

{
  "dossier_id": "<id>",
  "record_id": "7A-D-000731",
  "status": "ready",
  "sealed_at": "<time the seal was applied>",
  "download": "<expiring link>"
}

04Receive a flash

Sketch, not a live endpoint

A webhook body, sent to an address you register. The rule that raised the tier travels with it.

Request

POST <your-receiver>/hooks/7analytica
X-Signature: <scheme not yet defined>
Content-Type: application/json

Response

{
  "event": "flash.raised",
  "tier": 3,
  "rule": "Two tier 2 alerts within 60 minutes",
  "record_ids": ["7A-D-000731", "7A-D-000735"],
  "raised_at": "2026-09-14T04:40:00Z"
}

04Design principles

Rules we would hold it to.

They are 7Analytica’s own rules, applied to a machine reader.

Provenance travels with every record: No endpoint returns a score without its inputs, and none returns a claim without the count of origins behind it. A reading that arrives without its working is a bug.
Read-only first: The first thing we would build is reading. Writing back, such as notes or watchlists, comes later if people need it.
Rate limits are planned: We expect to limit requests so that one client cannot crowd out others. The limits are not set, and we will publish them before anyone depends on them.
Stable references: Records keep their permanent . A path that works today keeps meaning the same thing, and a breaking change gets a new version.
Describe, never conclude: The API carries the same rule as the rest of 7Analytica. It returns evidence and readings. It does not return a verdict or a recommendation.

05Shape the API and MCP server

Shape the API and MCP server.

Tell us what you would build. Two or three lines is enough: what you would pull, how often, and what it would feed. We read every reply, and the replies decide what we scope first.

Shape the API and MCP server Book a call

Prefer email? Write to bhuvanchandra@7analytica.com.

What you would read: the , audits, , alerts.
Whether you would use the API, the MCP server for an agent, or both.
How often, and how much: a daily export, or a stream.
Where it would land: a spreadsheet, a case system, a data warehouse.
What would stop you from using it: a format, a limit, a sign-in rule.

Replying does not reserve access, join a queue or commit either side. It tells us what to build first.

06Questions

Questions about the API and MCP server.

Is there a 7Analytica API or MCP server today?

No. A self-serve API and an MCP server are planned and not built. No endpoint, key, sandbox or beta is open, and the code samples on the developers page are sketches with placeholder hosts.

When will the API be available?

A self-serve API is planned and not built. The live data service it would sit on is in place. The roadmap shows when something moves from planned to built.

Can I get early access?

There is no early access list. You can tell us what you would build through the interest form. It helps us decide what to scope first, and it does not reserve access.

What would the API cover?

Reading the , a record's trust and bias readings and , export, webhooks for flashes and changes, and exports in JSON, CSV and PDF. Single sign-on with SAML or OIDC is a separate planned item.

What is the MCP server?

MCP is the Model Context Protocol, a common way for AI agents and developer tools to call outside services. A 7Analytica MCP server would let an agent read records and audits with their . It is planned, not built, and its tools are not defined yet.

Will the API return scores without their inputs?

No. The design principle is that travels with every record: a score comes with its inputs and a claim comes with the count of origins behind it.

Will it be read-only?

Read-only is the first thing we would build. Writing back would come later, if people need it.

Does it cost anything?

Not decided. No price is published for any part of 7Analytica, and access is arranged by request. See the pricing page.

See what is built today.

The wire, the record, the audit and the dossier all work, and you can walk through them with no sign-up.

Try the live demoRead the roadmap

An API and an MCP server are planned.

What this page describes

Planned, not built

What a call could look like.

About these sketches

What it could cover.

What exists, and what does not.

The wire, the record, the audit and the dossier, as screens

A live data service behind the wire

A self-serve API

An MCP server for agents

Webhooks and alerts

Single sign-on, SAML or OIDC

Programmatic access terms and pricing

What it could cover.

What a call could look like.

Rules we would hold it to.

Shape the API and MCP server.

Questions about the API and MCP server.

Is there a 7Analytica API or MCP server today?

When will the API be available?

Can I get early access?

What would the API cover?

What is the MCP server?

Will the API return scores without their inputs?

Will it be read-only?

Does it cost anything?

See what is built today.