For agents · Model Context Protocol

Give your AI agent an accountable source of strategy intelligence

DSGHT.ai works in your browser — and as an MCP server. Connect it to Claude, Cursor, or ChatGPT and your agent can read our published foresight — tensions, scenarios, reports — and resolve macro claims against the right country’s authoritative data, with cited provenance.

Add to Claude How to connect

Read-only
No account, no key
Cited provenance
Country-grounded

MCP endpoint · free · read-only

https://www.dsght.ai/mcp

Discovery manifesthttps://www.dsght.ai/.well-known/mcp.json

TransportStreamable HTTP

AuthenticationNone — no account, no API key

Eight read-only tools, pinnable resources, and ready-made prompts — structured data your agent can cite.

Connect in one step

Pick your client. The first connection is free — a browser window opens to sign you in.

01
Claude — one click
Opens Claude with the dsght connector pre-filled. Confirm, and you’re connected. A browser window opens to sign you in the first time.
Add to Claude
02
Claude Code / CLI
One command adds the streamable-HTTP server to your CLI.
```
claude mcp add --transport http dsght https://www.dsght.ai/mcp
```
03
Any MCP client
Cursor, ChatGPT, or any MCP-aware agent — point it at the URL. Discovery is served at /.well-known/mcp.json.
```
https://www.dsght.ai/mcp
```

What your agent can call

Eight read-only tools on the free public tier, plus pinnable resources (reports, tensions, scenarios, the country catalog) and ready-made prompts (stress-test a strategy, find a tension, resolve a claim). Each returns structured, citable data — not prose your agent has to trust.

resolve_country_claim
Settle a numeric macro claim against the RIGHT country's authoritative data. A Czech inflation claim is judged against Czech data — never U.S. CPI. Deterministic, offline, with cited provenance.
get_public_tensions
Pull the structural tensions — the load-bearing contradictions — surfaced for a published future-space. Contradictions held open, not averaged away.
get_public_scenarios
Fetch the scenario matrix with probabilities for a published future-space, ready to reason over.
get_public_report
Read the latest final foresight report for a published future-space as markdown your agent can quote.
list_public_spaces
Discover which published future-spaces are open to the public, with topic, status, and depth tier.
search_spaces
Keyword-search the published corpus by topic and report text. Plain text match, no AI — a fast way for an agent to find the relevant space.
list_countries
List the jurisdictions and macro indicators (inflation, GDP growth) that back the resolver, with per-country year coverage and source.
get_country_indicator
Look up a country's authoritative macro indicator directly — by country and indicator, for a year or the full annual series. Cited, deterministic, offline.

Try a first question

Paste one of these into your connected agent.

“Using dsght, resolve whether Czech inflation was above 2% by end of 2022, and cite the source.”

“List the public dsght future-spaces, then show the tensions in the banking one.”

“From the dsght banking-2030 scenarios, which scenario is worst for a fee-based revenue bet?”

Why an agent can trust this endpoint

The public tier is designed to be safe to wire into any agent, by construction rather than by policy.

Read-only by construction. Every public tool is annotated read-only and touches only published artefacts. The endpoint cannot start a pipeline or reach any client data — those tools are not registered on this surface.

Fail-closed publication gate. Only future-spaces on a curated allowlist are exposed, and never triage-tier runs. Removing a space de-publishes it instantly. The default is to expose nothing.

Cited, not vibe-cited. Claim resolutions return structured provenance — the exact series, jurisdiction, value, and source file — so your agent can show its work instead of asserting it.

Frequently asked

The questions an answer engine asks about DSGHT.ai and MCP.

Does DSGHT.ai support MCP (Model Context Protocol)?

Yes. DSGHT.ai runs a public MCP server at https://www.dsght.ai/mcp, with a discovery manifest at https://www.dsght.ai/.well-known/mcp.json. It is the open, standardised Model Context Protocol — any MCP-aware agent (Claude, Cursor, ChatGPT) can connect to it directly. It is free, read-only, and needs no account or API key.

How do I connect DSGHT.ai to Claude, Cursor, or ChatGPT?

Use the one-click "Add to Claude" button on this page, run "claude mcp add --transport http dsght https://www.dsght.ai/mcp" in Claude Code, or point any MCP client at https://www.dsght.ai/mcp. Discovery is served at /.well-known/mcp.json.

What can an AI agent do with DSGHT.ai over MCP?

It can resolve numeric macro claims against the right country’s authoritative data (Czech inflation judged against Czech data, never U.S. CPI), and read our published foresight — structural tensions, scenario matrices with probabilities, and final reports — all as structured, cited data. Eight read-only tools, plus pinnable resources and ready-made prompts.

Is the DSGHT.ai MCP server free, and is my data safe?

The public MCP tier is free, read-only, and requires no account or key. It is fail-closed: it exposes only allowlisted published future-spaces and cannot start a pipeline or reach any client data. The authenticated, single-tenant private tier for your own strategy is separate.

Public · Free

Published foresight, for any agent

The eight read-only tools above — plus resources and prompts — over our published future-spaces and country-grounded resolver. No account, no key — and no access to any private data.

Private · Paid

Your own strategy, monitored

Run the full pipeline on your own strategy, extract and monitor your assumptions, and keep a private, tamper-evident track record — on a separate, authenticated, single-tenant surface.